Assessing real-time Zika risk in the United States

Abstract

Background
Confirmed local transmission of Zika Virus (ZIKV) in Texas and Florida have heightened the need for early and accurate indicators of self-sustaining transmission in high risk areas across the southern United States. Given ZIKV’s low reporting rates and the geographic variability in suitable conditions, a cluster of reported cases may reflect diverse scenarios, ranging from independent introductions to a self-sustaining local epidemic.

Methods
We present a quantitative framework for real-time ZIKV risk assessment that captures uncertainty in case reporting, importations, and vector-human transmission dynamics.

Results
We assessed county-level risk throughout Texas, as of summer 2016, and found that importation risk was concentrated in large metropolitan regions, while sustained ZIKV transmission risk is concentrated in the southeastern counties including the Houston metropolitan region and the Texas-Mexico border (where the sole autochthonous cases have occurred in 2016). We found that counties most likely to detect cases are not necessarily the most likely to experience epidemics, and used our framework to identify triggers to signal the start of an epidemic based on a policymakers propensity for risk.

Conclusions
This framework can inform the strategic timing and spatial allocation of public health resources to combat ZIKV throughout the US, and highlights the need to develop methods to obtain reliable estimates of key epidemiological parameters.

Electronic supplementary material
The online version of this article (doi:10.1186/s12879-017-2394-9) contains supplementary material, which is available to authorized users.

Full text

R E S E A R C H A R T I C L E Open Access
Assessing real-time Zika risk in the United
States
Lauren A. Castro
1†
, Spencer J. Fox
1*†
, Xi Chen
2
, Kai Liu
3
, Steven E. Bellan
4,5
, Nedialko B. Dimitrov
2
,
Alison P. Galvani
6,7
and Lauren Ancel Meyers
1,8
Abstract
Background: Confirmed local transmission of Zika Virus (ZIKV) in Texas and Florida have heightened the need for
early and accurate indicators of self-sustaining transmission in high risk areas across the southern United States.
Given ZIKV’s low reporting rates and the geographic variability in suitable conditions, a cluster of reported cases
may reflect diverse scenarios, ranging from independent introductions to a self-sustaining local epidemic.
Methods: We present a quantitative framework for real-time ZIKV risk assessment that captures uncertainty in case
reporting, importations, and vector-human transmission dynamics.
Results: We assessed county-level risk throughout Texas, as of summer 2016, and found that importation risk was
concentrated in large metropolitan regions, while sustained ZIKV transmission risk is concentrated in the
southeastern counties including the Houston metropolitan region and the Texas-Mexico border (where the sole
autochthonous cases have occurred in 2016). We found that counties most likely to detect cases are not necessarily
the most likely to experience epidemics, and used our framework to identify triggers to signal the start of an
epidemic based on a policymakers propensity for risk.
Conclusions: This framework can inform the strategic timing and spatial allocation of public health resources to
combat ZIKV throughout the US, and highlights the need to develop methods to obtain reliable estimates of key
epidemiological parameters.
Keywords: Zika, ZIKV, Importation risk, Autochthonous transmission risk
Background
In February 2016, the World Health Organization (WHO)
declared Zika virus (ZIKV) a Public Health Emergency of
International Concern [1]. Though the Public Health
Emergency has been lifted, ZIKV still poses a great threat
for reemergence in susceptible regions in seasons to come
[2]. In the US, the 268 reported mosquito-borne autoch-
thonous (local) ZIKV cases occurred in Southern Florida
and Texas, with the potential range of a primary ZIKV
vector, Aedes aegypti, including over 30 states [3–5]. Of
the 2487 identified imported ZIKV cases in the US
through the end of August, 137 had occurred in Texas.
Given historic small, autochthonous outbreaks (ranging
from 4 to 25 confirmed cases) of another arbovirus
vectored by Ae. Aegypti—dengue (DENV) [5–7], Texas
was known to be at risk for autochthonous arbovirus
transmission, and the recent outbreaks have highlighted
the need for increased surveillance and optimized re-
source allocation in the states and the rest of the vulner-
able regions of the Southern United States.
As additional ZIKV waves are possible in summer
2017, public health professionals will continue to face
considerable uncertainty in gauging the severity, geo-
graphic range of local outbreaks, and appropriate timing
of interventions, given the large fraction of undetected
ZIKV cases (asymptomatic) and economic tradeoffs of
disease prevention and response [8–11]. Depending on
the ZIKV symptomatic fraction, reliability and rapidity
of diagnostics, importation rate, and transmission rate,
the detection of five autochthonous cases in a Texas
county, for example, may indicate a small chain of cases
* Correspondence: spncrfx@gmail.com
†
Equal contributors
1
Department of Integrative Biology, The University of Texas at Austin, Austin,
TX, USA
Full list of author information is available at the end of the article
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Castro et al. BMC Infectious Diseases (2017) 17:284
DOI 10.1186/s12879-017-2394-9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

from a single importation, a self-limiting outbreak, or a
large, hidden epidemic underway (Fig. 1). These diver-
ging possibilities have precedents. In French Polynesia, a
handful of ZIKV cases were reported by October 2013; 2
months later an estimated 14,000–29,000 individuals
had been infected [8, 9]. By contrast, Anguilla had 17
confirmed cases from late 2015 into 2016 without a sub-
sequent epidemic, despite large ZIKV epidemics in sur-
rounding countries [12]. To address the uncertainty, the
CDC issued guidelines for state and local agencies; they
recommend initiation of public health responses follow-
ing local reporting of two non-familial autochthonous
ZIKV cases [13].
Previous risk assessments of ZIKV have provided static
a priori assessments based on historical incidence and
vector suitability, but they do not provide dynamic risk
assessments as cases accumulate in a region. Here, we
present a framework to support real-time risk assess-
ment, and demonstrate its application in Texas. Our
framework accounts for the uncertainty regarding ZIKV
epidemiology, including importation rates, reporting
rates, local vector populations, and socioeconomic con-
ditions, and can be readily updated as our understanding
of ZIKV evolves. To estimate current and future epi-
demic risk from real-time ZIKV case reports, the model
incorporates a previously published method for estimat-
ing local ZIKV transmission risk and a new model for
estimating local importation risk. Across Texas’254
counties, we find that the estimated risk of a locally sus-
tained ZIKV outbreak rises precipitously as autochthon-
ous cases accumulate, and that counties at the southern
tip of the Texas-Mexico border and in the Houston
Metropolitan Area are at the highest risk for ZIKV
transmission. This statewide variation in risk stems pri-
marily from mosquito suitability and socio-environmental
constraints on ZIKV transmission rather than heterogen-
eity in importation rates.
Methods
Our risk-assessment framework is divided into three sec-
tions: (1) county-level epidemiological estimates of ZIKV
importation and relative transmission rates, (2) county-
specific ZIKV outbreak simulations, and (3) ZIKV risk
analysis (Additional file 1: Figure S1). To demonstrate this
approach, we estimate county-level ZIKV risks throughout
the state of Texas for August 2016, given that, by May
2016, Texas experienced dozens of ZIKV importations
without subsequent vector-borne transmission.
Estimating importation rates
Our analysis assumes that any ZIKV outbreaks in Texas
originate with infected travelers returning from active
ZIKV regions. To estimate the ZIKV importation rate for
specific counties, we (1) estimated the Texas statewide im-
portation rate (expected number of imported cases per
day) for August 2016, (2) estimated the probability (import
risk) that the next Texas import will arrive in each county,
and (3) took the product of the state importation rate and
each county importation probability.
1. During the first quarter of 2016, 27 ZIKV travel-
associated cases were reported in Texas [5], yielding
a baseline first quarter estimate of 0.3 imported
cases/day throughout Texas. In 2014 and 2015,
arbovirus introductions into Texas increased
threefold over this same time period, perhaps
0
5
10
0 25 50 75 100
Time (days)
Prevalence
Asymptomatic
Symptomatic
0
5
10
0 25 50 75 100
Time (days)
Prevalence
A
B
Detection
Importation
Fig. 1 ZIKV emergence scenarios. A ZIKV infection could spark (a) a self-limiting outbreak or (b) a growing epidemic. Cases are partitioned into
symptomatic (grey) and asymptomatic (black). Arrows indicate new ZIKV importations by infected travelers and vertical dashed lines indicate case
reporting events. On the 75th day, these divergent scenarios are almost indistinguishable to public health surveillance, as exactly three cases have
been detected in both. By the 100th day, the outbreak (a) has died out with 21 total infections while the epidemic (b) continues to grow with
already 67 total infections. Each scenario is a single stochastic realization of the model with R
0
= 1.1, reporting rate of 10%, and introduction rate
of 0.1 case/day
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 2 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

driven by seasonal increases in arbovirus activity in
endemic regions and the approximately 40% increase
from quarter 1 to quarter 3 in international travelers
to the US [14]. Taking this as a baseline (lower
bound) scenario, we projected a corresponding
increase in ZIKV importations to 0.9 cases/day
(statewide) for the third quarter.
2. To build a predictive model for import risk, we fit a
probabilistic model (maximum entropy) [15]of
importation risk to 183 DENV, 38 CHIKV, and 31
ZIKV Texas county-level reported importations from
2002 to 2016 and 10 informative socioeconomic,
environmental, and travel variables (Additional file 1:
section 1.1). Given the geographic and biological
overlap between ZIKV, DENV and Chikungunya
(CHIKV), we used historical DENV and CHIKV
importation data to supplement ZIKV importations
in the importation risk model, while recognizing that
future ZIKV importations may be fueled by large
epidemic waves in neighboring regions and summer
travel, and thus far exceed recent DENV and CHIKV
importations [16]. Currently, DENV, CHIKV, and
ZIKV importation patterns differ most noticeably
along the Texas-Mexico border. Endemic DENV
transmission and sporadic CHIKV outbreaks in
Mexico historically have spilled over into neighboring
Texas counties. In contrast, ZIKV is not yet as
widespread in Mexico as it is in Central and
South America, with less than 10 reported ZIKV
importations along the border to date (October 2016).
We included DENV and CHIKV importation data in
the model fitting so as to consider potential future
importation pressure from Mexico, as ZIKV continues
its increasing trend since March 2016 [17]. To find
informative predictors for ZIKV importation risk,
we analyzed 72 socio-economic, environmental,
and travel variables, and removed near duplicate
variables and those that contributed least to model
performance, based on out-of-sample cross validation
of training and testing sets of data [18,19], reducing
the original set of 72 variables to 10 (Additional file 1:
Tables S3-S4). We validated our importation model
by comparing the predicted distribution of cases
across the state given a total number of imported
cases (September 2016) as a linear predictor of the
empirical distribution of cases across counties.
County transmission rates (R
0
)
The risk of ZIKV emergence following an imported case
will depend on the likelihood of mosquito-borne transmis-
sion. For emerging diseases like ZIKV, the public health
and research communities initially face considerable uncer-
tainty in the drivers and rates of transmission, given the
lack of field and experimental studies and epidemiological
data, and often derive insights through analogy to similar
diseases. For our case study, we estimated county-level
ZIKV transmission potential by Ae. aegypti using a recently
published model [20], that derives some of its key parame-
ters from DENV data. The utility of our framework de-
pends on the validity of such estimates and will increase as
our knowledge of ZIKV improves. However, we expect our
results to be robust to most sources of uncertainty regard-
ing ZIKV and DENV epidemiology, as they may influence
the absolute but not relative county-level risks.
We estimated the ZIKV reproduction number (R
0
),
the average number of secondary infections caused by a
single infectious individual in a fully susceptible popula-
tion, for each Texas county following the method de-
scribed in Perkins et al. [20]. The method calculates R
0
using a temperature-dependent formulation of the Ross-
Macdonald model, where mosquito mortality rate (μ)and
extrinsic incubation period of ZIKV (n) are temperature
dependent functions; the human-mosquito transmission
probability (b = 0.4), number of days of human infectious-
ness (c/r= 3.5), and the mosquito biting rate (a = 0.67)
are held constant at previously calculated values [20–25];
and the economic-modulated mosquito-human contact
scaling factor (m) is a function of county mosquito abun-
dance and GDP data fit to historic ZIKV seroprevalence
data [20]. To account for uncertainty in the temperature-
dependent functions (the extrinsic incubation period (EIP)
and mosquito mortality rate) and in the relationship be-
tween economic index and the mosquito-to-human con-
tact rate, Perkins et al. generated functional distributions
via 1000 Monte Carlo samples from the underlying par-
ameter distributions. We assume DENV estimates for
these temperature-dependent functions, since we lack
such data for ZIKV and these Flaviviruses are likely to ex-
hibit similar relationships between temperature and EIP in
Ae. Aegypti [25]. We used the resulting distributions to es-
timate R
0
for each county, based on county estimates for
the average August temperature, mosquito abundance
from Kraemer et al. [24], and GDP [25]. Our R
0
estimates
were similar to those reported by Perkins et al. [20]
with 95% confidence intervals spanning from 0 to 3.1
(Additional file 1: Figure S3). Given this uncertainty, and
that our primary aim is to demonstrate the risk assessment
framework rather than provide accurate estimates of R
0
for
Texas, we use these estimates to estimate relative county-
level transmission risks (by scaling the county R
0
estimates
from 0 to 1). In each simulation, we assume that a county’s
R
0
is the product of its relative risk and a chosen maximum
R
0
.Forourcasestudy,weassumeamaximumcounty-level
R
0
of 1.5 This is consistent with historical arbovirus activity
in Texas (which has never sustained a large arbovirus epi-
demic) and demonstrates the particular utility of the ap-
proach in distinguishing outbreaks from epidemics around
the epidemic threshold of R
0
=1.
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 3 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

ZIKV outbreak simulation model
Assuming mosquito-borne transmission as the main
driver of epidemic dynamics, to transmit ZIKV, a mos-
quito must bite an infected human, the mosquito must
get infected with the virus, and then the infected mos-
quito must bite a susceptible human. Rather than expli-
citly model the full transmission cycle, we aggregated
the two-part cycle of ZIKV transmission (mosquito-to-
human and human-to-mosquito) into a single exposure
period where the individual has been infected by ZIKV,
but not yet infectious, and do not explicitly model
mosquitos. For the purposes of this study, we need only
ensure that the model produces a realistic human-to-
human generation time of ZIKV transmission, and the
simpler model is more flexible to disease transmission
pathways. We fit the generation time of the ZIKV model
to early ZIKV Epidemiological estimates, with further fit-
ting details described in Additional file 1: section 2.4.
The resultant model thus follows a Susceptible-Exposed-
Infectious-Recovered (SEIR) transmission process stem-
ming from a single ZIKV infection using a Markov
branching process model (Additional file 1: Figure S4). The
temporal evolution of the compartments is governed by
daily probabilities of infected individuals transitioning be-
tween disease states. New cases arise from importations or
autochthonous transmission (Additional file 1: Table S5).
We treat days as discrete time steps, and the next disease
state progression depends solely on the current state and
thetransitionprobabilities.Weassumethatinfectious
cases cause a Poisson distributed number of secondary
cases per day (via human to mosquito to human transmis-
sion), but this assumption can be relaxed as more informa-
tion regarding the distribution of secondary cases becomes
available. We also assume infectious individuals are intro-
duced daily according to a Poisson distributed number of
cases around the importation rate. Furthermore, Infectious
cases are categorized into reported and unreported
cases according to a reporting rate. We assume that
reporting rates approximately correspond to the per-
centage (~20%) of symptomatic ZIKV infections [10]
and occur at the same rate for imported and locally ac-
quired cases. Additionally, we make the simplifying as-
sumption that reported cases transmit ZIKV at the same
rate as unreported cases. We track imported and autoch-
thonous cases separately, and conduct risk analyses based
on reported autochthonous cases only, under the assump-
tion that public health officials will have immediate and
reliable travel histories for all reported cases [13].
Simulations
For each county risk scenario, defined by an importation
rate, transmission rate, and reporting rate, we ran 10,000
stochastic simulations. Each simulation began with one
imported infectious case and terminated either when
there were no individuals in either the Exposed or Infec-
tious classes or the cumulative number of autochthon-
ous infections reached 2000. Thus the total outbreak
time may differ across simulations. We held R
0
constant
throughout each simulation, as we sought to model early
outbreak dynamics over short periods (relative to the
seasonality of transmission) following introduction. We
classified simulations as either epidemics or self-limiting
outbreaks; epidemics were simulations that fulfilled two
criteria: reached 2000 cumulative autochthonous infec-
tions and had a maximum daily prevalence (defined as
the number of current infectious cases) exceeding 50 au-
tochthonous cases (Additional file 1: Figure S6). The sec-
ond criterion distinguishes simulations resulting in large
self-sustaining outbreaks (that achieve substantial peaks)
from those that accumulate infections through a series of
small, independent clusters (that fail to reach the daily
prevalence threshold). The latter occurs occasionally
under low R
0
s and high importation rates scenarios.
To verify that our simulations do not aggregate cases
from clear temporally separate clusters, we calculated
the distribution of times between sequential cases
(Additional file 1: Figure S7). In our simulated epi-
demics, almost all sequentially occurring cases occur
within 14 days of each other, consistent with the
CDC’s threshold for identifying local transmission
events (based on the estimated maximum duration of
the ZIKV incubation period) [13].
Outbreak analysis
Our stochastic framework allows us to provide multiple
forms of real-time county-level risk assessments as re-
ported cases accumulate. For each county, we found the
probability that an outbreak will progress into an epi-
demic, as defined above, as a function of the number of
reported autochthonous cases. We call this epidemic
risk. To solve for epidemic risk in a county following the
xth reported autochthonous case, we first find all simu-
lations that experience at least xreported autochthonous
cases, and then calculate the proportion of those that
are ultimately classified as epidemics. For example, con-
sider a county in which 1000 of 10,000 simulated out-
breaks reach at least two reported autochthonous cases
and only 50 of the 1000 simulations ultimately fulfill the
two epidemic criteria; the probability of detecting two
cases in the county would be 10% and the estimated epi-
demic risk following two reported cases in that county
would be 5%. This simple epidemic classification scheme
rarely misclassifies a string of small outbreaks as an epi-
demic, with the probability of such an error increasing
with the importation rate. For example, epidemics should
not occur when R
0
= 0.9. If the importation rate is high,
overlapping series of moderate outbreaks occasionally
meet the two epidemic criteria. Under the highest
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 4 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

importation rate we considered (0.3 cases/day), only 1% of
outbreaks were misclassified.
This method can be applied to evaluate universal trig-
gers (like the recommended two-case trigger) or derive ro-
bust triggers based on risk tolerance of public health
agencies. For example, if a policymaker would like to initi-
ate interventions as soon as the risk of an epidemic
reaches 30%, we would simulate local ZIKV transmission
and solve for the number of reported cases at which the
probability of an epidemic first exceeds 30%. Generally,
the recommended triggers decrease (fewer reported cases)
as the policymaker threshold for action decreases, (e.g.
10% versus 30% threshold) and as the local transmission
potential increases (e.g. R
0
= 1.5 versus R
0
=1.2).
Results
ZIKV importation risk within Texas is predicted by vari-
ables reflecting urbanization, mobility patterns, and socio-
economic status (Additional file 1: Table S3), and is
concentrated in metropolitan counties of Texas (Fig. 2a). In
comparing the predictions of this model to out-of-sample
data from April to September 2016, the model underesti-
mated the statewide total number of importations (81 vs
151), but robustly predicted the relative importation rates
between counties (β= 0.97, R
2
= 0.74, p<0.001).Thetwo
highest risk counties–Harris, which includes Houston, and
Travis, which includes Austin, have an estimated 27% and
10% chance of receiving the next imported Texas case re-
spectively and contain international airports.
ZIKV transmission risk is concentrated in southeastern
Texas (Fig. 2b), partially overlapping with regions of high
importationrisk(Fig.2a).Ourcounty-levelestimatesofR
0
range widely (from 0.8 to 3.1 for the highest-risk county),
reflecting the uncertainty in socioeconomic and environ-
mental drivers of ZIKV (Additional file 1: Figure S3). We
therefore analyzed the relative rather than absolute trans-
mission risks. For purposes of demonstration, we assumed
a plausible maximum county-level R
0
of 1.5, which closely
followed our median estimates, and scaled the transmission
risk for each county accordingly. The following risk ana-
lyses can be readily refined as we gain more precise and lo-
calized estimates of ZIKA transmission potential.
Wide ranges of outbreaks are possible under a single set
of epidemiological conditions (Fig. 3a). The relationship be-
tween what policymakers can observe (cumulative reported
cases) and what they wish to know (current underlying dis-
ease prevalence) can be obscured by such uncertainty, and
will depend critically on reporting rates (Fig. 3b). Under a
scenario estimated for Cameron County which experienced
the only autochthonous ZIKV transmission in Texas and
with a 20% reporting rate, ten linked and reported autoch-
thonous cases correspond to 6 currently circulating cases
with a 95% CI of 1–16 from inherent, early-stage outbreak
stochasticity. From this wide range of outbreak trajectories,
we can characterize time-varying epidemic risk as cases ac-
cumulate in a given county. We track the probability of
epidemic expansion following each additional reported case
in high and low reporting rate scenarios (Fig. 3c).
Dallas
Houston
San Antonio
Austin
El Paso
McAllen
Corpus Christi
Brownsville
Kileen
Beaumont
0.002 0.02 0.2
Import Probability
A
0.25 0.50 0.75 1.00
Relative Transmission Risk
B
Fig. 2 ZIKV importation and transmission risk estimates across Texas for August 2016. aColor indicates the probability that the next ZIKV import
will occur in a given county for each of the 254 Texas counties. Probability is colored on a log scale. The 10 most populous cities in Texas are
labeled. Houston’s Harris County has 2.7 times greater chance than Austin’s Travis County of receiving the next imported case. bEstimated
county-level transmission risk for ZIKV (See Additional file 1: Figure S7 for seasonal differences). Harris county and Dallas County rank among the
top 5 and top 10 for both importation and transmission risk respectively; counties in McAllen and Houston metropolitan area rank among the
top 20. Bolded county border indicates counties with recorded local ZIKV transmission
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 5 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

These curves can support both real-time risk assess-
ment as cases accumulate and the identification of sur-
veillance triggers indicating when risk exceeds a
specified threshold. For example, suppose a policymaker
wanted to initiate an intervention upon two reported
cases, this would correspond with a 49% probability of
an epidemic if 10% of cases are reported, but only 25% if
the reporting rate is doubled. Alternatively suppose a
policy maker wishes to initiate an intervention when the
chance of an epidemic exceeds 50%. In the low reporting
rate scenario, they should act immediately following the
third autochthonous reported case, but could wait until
the eleventh case with the high reporting rate.
To evaluate a universal intervention trigger of two re-
ported autochthonous cases, we estimate both the prob-
ability of two reported cases in each county and the level
of epidemic risk at the moment the trigger event occurs
(second case reported). Assuming a baseline importation
rate extrapolated from importation levels in March 2016
to August 2016, county R
0
scaled from a maximum of
1.5, and a 20% reporting rate, only a minority of counties
are likely to experience a trigger event (Fig. 4a). While
247 of the 254 counties (97%) have non-zero probabil-
ities of experiencing two reported autochthonous cases,
only 86 counties have at least a 10% chance of such an
event (assuming they experience at least one import-
ation), with the remaining 168 counties having a median
probability of 0.0038 (range 0.0005 to 0.087). Assuming
that a second autochthonous case has indeed been re-
ported, we find that the underlying epidemic risk varies
widely among the 247 counties, with most counties hav-
ing near zero epidemic probabilities and a few counties
far exceeding a 50% chance of epidemic expansion. For
example, two reported autochthonous cases in Harris
County, correspond to a 99% chance of ongoing trans-
mission that would proceed to epidemic proportions
without intervention, with the rest of the Houston
metropolitan also at relatively high risk ranging from 0
(Galveston) to 90% (Waller) (Fig. 4b).
Given that a universal trigger may signal disparate
levels of ZIKV risk, policy makers might seek to adapt
their triggers to local conditions. Suppose a policymaker
wishes to design triggers that indicate a 50% chance of
an emerging epidemic (Fig. 4c). Under the baseline im-
portation and reporting rates, an estimated 31 of the 254
counties in Texas are expected to reach a 50% epidemic
probability, with triggers ranging from one (Harris
County) to 21 (Jefferson County) reported autochthon-
ous cases, with a median of two cases. Counties who de-
tect cases simply due to high importation rates do not
have triggers, and the magnitude of a trigger helps quan-
tify a county’s absolute risk for an epidemic as a function
of the reported autochthonous cases.
Discussion
Our framework provides a data-driven approach to esti-
mating ZIKA emergence risks from potentially sparse
and biased surveillance data [26, 27]. By mapping ob-
served cases to current and future risks, in the face of
considerable uncertainty, the approach can also be used
to design public health action plans and evaluate the
utility of local versus regional triggers. We demonstrate
its application across the 254 ecologically and demo-
graphically diverse counties of Texas, one of the two
0
10
20
30
40
50
0 50 100 150
Time (days)
Reported Autochthonous Cases
5
10
20
0 5 10 15
Reported Cases
Cases (log scale)
Reporting Rate
10%
20%
0.00
0.25
0.50
0.75
1.00
0 5 10 15
Reported Cases
Epidemic Probability
ABC
Fig. 3 Real-time risk-assessment for ZIKV transmission. All figures are based on transmission and importation risks estimated for Cameron County,
Texas. aTwo thousand simulated outbreaks. bTotal number of (current) autochthonous cases as a function of the cumulative reported autochthonous
cases, under a relatively high (dashed) or low (solid) reporting rate. Ribbons indicate 50% quantiles. cThe increasing probability of imminent epidemic
expansion as reported autochthonous cases accumulate for a low (solid) and high (dashed) reporting rate. Suppose a policy-maker plans to trigger a
public health response as soon as a second case is reported (vertical line). Under a 10% reporting rate, this trigger would correspond to a 49%
probability of an ensuing epidemic. Under a 20% reporting rate, the probability would be 25%
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 6 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

states that has sustained autochthonous ZIKV outbreaks
[6, 7]. The approach requires local estimates of ZIKV
importation and transmission rates. For the Texas ana-
lysis, we developed a novel model for estimating county-
level ZIKV importation risk and applied published
methods to estimate relative county-level transmission
risks (Fig. 2). We expect that most Texas counties are
not at risk for a sustained ZIKV epidemic (Fig. 4), and
find that many of the highest risk counties lie in the
southeastern region surrounding the Houston metropol-
itan area and the lower Rio Grande valley. However, R
0
estimates are uncertain, leaving the possibility that the
R
0
could be as high as other high risk regions that sus-
tained epidemics [20, 28, 29]. Our analysis is consistent
with historic DENV and CHIKV outbreaks and correctly
identifies Cameron county, the only Texas county to
have reported local transmission, as a potential ZIKV
hot-spot, especially when November estimates are used
[30] (Additional file 1: Figure S9).
Surveillance triggers–guidelines specifying situations
that warrant intervention–are a key component of many
public health response plans. Given the urgency and un-
certainty surrounding ZIKV, universal recommendations
can be both pragmatic and judicious. To assist Texas
policymakers in interpreting the two-case trigger for
intervention guidelines issued by the CDC [13], we used
our framework to integrate importation and transmis-
sion risks and assess the likelihood and implication of a
two-case event for each of Texas’254 counties, under a
scenario projected from recent ZIKV data to August
2016. Across counties, there is enormous variation in
both the chance of a trigger and the magnitude of the
public health threat if and when two cases are reported.
Given this variation, rather than implement a universal
trigger, which may correspond to different threats in dif-
ferent locations, one could design local surveillance trig-
gers that correspond to a universal risk threshold. Our
modeling framework can readily identify triggers (num-
bers of reported cases) for indicating any specified epi-
demic event (e.g., prevalence reaching a threshold or
imminent epidemic expansion) with any specified risk
tolerance (e.g., 10% or 50% chance of that the event has
occurred), given local epidemiological conditions. We
found close agreement between the recommended two-
case trigger and our epidemic derived triggers based on
a 50% probability of expansion. Of the 30 counties with
derived triggers, the median trigger was 2, ranging from
one to 21 reported autochthonous cases. These findings
apply only to the early, pre-epidemic phase of ZIKV in
Texas, when importations occur primarily via travel
from affected regions outside the contiguous US.
These analyses highlight critical gaps in our under-
standing of ZIKV biology and epidemiology. The relative
transmission risks among Texas counties appear fairly
robust to these uncertainties, allowing us to identify high
risk regions, including Cameron County in the Lower
Rio Grande Valley. Public health agencies might there-
fore prioritize such counties for surveillance and inter-
ventions resources. Given the minimal incursions of
DENV and CHIKV into Texas over that past eleven
years since the first DENV outbreak in Cameron County,
and the high number of importations into putative hot-
spot counties without autochthonous transmission, we
suspect that, if anything, we may be underestimating the
socioeconomic and behavioral impediments to ZIKV
transmission in the contiguous US. Our analysis also
Dallas
Houston
San Antonio
Austin
0.00 0.25 0.50 0.75 1.00
Two−case Probability
A
Dallas
Houston
San Antonio
Austin
0.00 0.25 0.50 0.75 1.00
Epidemic Probability
B
Dallas
Houston
San Antonio
Austin
5 101520
Trigger (Reported Cases)
C
Fig. 4 Texas county ZIKV risk assessment. aProbability of an outbreak with at least two reported autochthonous ZIKV cases. bThe probability of
epidemic expansion at the moment the second autochthonous ZIKV case is reported in a county. White counties never reach two reported cases
across all 10,000 simulated outbreaks; light gray counties reach two cases, but never experience epidemics. cRecommended county-level surveillance
triggers (number of reported autochthonous cases) indicating that the probability of epidemic expansion has exceeded 50%. White counties indicate that
fewer than 1% of the 10,000 simulated outbreaks reached two reported cases.Allthreemapsassumea20%reportingrateandabaselineimportation
scenario for August 2016 (81 cases statewide per 90 days) projected from historical arbovirus data.
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 7 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

reveals the significant impact of the reporting rate on the
timeliness and precision of detection. If only a small frac-
tion of cases are reported, the first few reported cases may
correspond to an isolated introduction or a growing epi-
demic. In contrast, if most cases are reported, policy-
makers can wait longer for cases to accumulate to trigger
interventions and have more confidence in their epi-
demiological assessments. ZIKV reporting rates are ex-
pected to remain low, because an estimated 80% of
infections are asymptomatic, and DENV reporting rates
have historically matched its asymptomatic proportion
[10, 31]. Obtaining a realistic estimate of the ZIKV report-
ing rate is arguably as important as increasing the rate it-
self, with respect to reliable situational awareness and
forecasting. An estimated 8–22% of ZIKV infections were
reported during the 2013–2014 outbreak in French
Polynesia [29]; however estimates ranging from 1 to 10%
have been reported during the ongoing epidemic in
Columbia [2, 28]. While these provide a baseline estimate
for the US, there are many factors that could increase (or
decrease) the reporting rate, such as ZIKV awareness
among both the public and health-care practitioners, or
active surveillance of regions with recent ZIKV cases. Our
analysis assumes that all counties have the same case de-
tection probabilities. However, only 40 of the 254 Texas
counties maintain active mosquito surveillance and con-
trol programs, potentially leading to differences in case
detection rates and surveillance efficacy throughout the
state [32]. Thus, rapid estimation of the reporting rate
using both traditional epidemiological data and new viral
sequenced based methods [33] should be a high priority
as they become available.
Conclusions
Our framework can support the development of re-
sponse plans, by forcing policymakers to be explicit
about risk tolerance, that is, the certainty needed before
sounding an alarm, and quantifying the consequences of
premature or delayed interventions. For example, should
ZIKV-related pregnancy advisories be issued when there
is only 5% chance of an impending epidemic? 10%
chance? 80%? A policymaker has to weigh the costs of
false positives–resulting in unnecessary fear and/or
intervention–and false negatives–resulting in suboptimal
disease control and prevention–complicated by the diffi-
culty inherent in distinguishing a false positive from a
successful intervention. The more risk averse the policy-
maker (with respect to false negatives), the earlier the
trigger should be, which can be exacerbated by low
reporting rates, high importation rate, and inherent
ZIKV transmission potential. In ZIKV prone regions
with low reporting rates, even risk tolerant policymakers
should act quickly upon seeing initial cases; in lower risk
regions, longer waiting periods may be prudent.
Additional files
Additional file 1: Supplemental Information: A framework for assessing
real-time Zika risk in the United States. (PDF 10870 kb)
Additional file 2: Supplemental Data: Raw historic arbovirus Texas
county importation data. (XLS 9 kb)
Additional file 3: Supplemental Data: Texas county importation
predictor raw data. (XLS 81 kb)
Abbreviations
CHIKV: Chikungunya Virus; DENV: Dengue Virus; SEIR model: Susceptible-
Exposed-Infectious-Recovered epidemiological model; WHO: World Health
Organization; ZIKV: Zika virus
Acknowledgements
We acknowledge the Texas Department of State Health Services (DSHS) for
providing historic arbovirus county importation data, and MUG Kraemer and
A Perkins for providing Texas mosquito abundance data and technical
guidance. We also thank M Johansson and M Meltzer of the Centers for
Disease Control and Prevention (CDC) and Dr. John Hellerstedt,
commissioner of the Texas DSHS, for critical conversations regarding the
modeling framework and its application to Texas and national ZIKV planning
efforts. We finally acknowledge the Texas Advanced Computing Center
(TACC) at The University of Texas at Austin for providing High performance
computing resources that have contributed to the research results reported
within this paper. URL: http://www.tacc.utexas.edu.
Funding
LAC, SEB, LAM, and APG were supported by a National Institute of General
Medical Sciences MIDAS grant to LAM and APG (U01GM087719), and APG
received additional NIGMS support (U01GM105627). SJF, KL, and LAM were
funded through a grant from the Defense Threat Reduction Agency to LAM
(HDTRA-14-C-0114). SEB was additionally supported by the International
Clinics on Infectious Disease Dynamics and Data (ICI3D) program, which is
funded by a NIGMS grant (R25GM102149). NBD and XC were partially
supported by a contract with the Texas Department of State Health Services
(2015–047259-001). The funders did not contribute to the design, of the
study, analysis, or interpretation of data.
Availability of data and materials
The datasets generated and analyzed during the current study are available
in the Additional files 2 and 3: supplemental materials and the R package,
rtZIKVrisk, which can be accessed here: https://github.com/sjfox/rtZIKVrisk.
There was no individualized patient or medical data used in our study, and
all ZIKV case data is publicly available.
Authors’contributions
ND,APG,LAM,LAC,SEB,andSJFdevelopedtheconceptualmodelingframework
and study design. LAC, SJF, and XC collected the data used in the analysis. LAC,
KL, and SJF reviewed relevant literature. XC and ND developed and completed
the Texas importation risk analysis. LAM, LAC, and SJF interpreted all results. LAC
and SJF performed the analyses, created the figures, and wrote the first draft. All
authors contributed to the presentation of the results, and the writing and
approval of the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Consent for publication
Not applicable.
Ethics approval and consent to participate
Not applicable.
Publisher’sNote
Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 8 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Author details
1
Department of Integrative Biology, The University of Texas at Austin, Austin,
TX, USA.
2
Graduate Program in Operations Research and Industrial
Engineering, The University of Texas at Austin, Austin, TX, USA.
3
Institute for
Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX,
USA.
4
Center for Ecology of Infectious Diseases, University of Georgia, Athens,
GA, USA.
5
Department of Epidemiology and Biostatistics, University of
Georgia, Athens, Athens, GA, USA.
6
Center for Infectious Disease Modeling
and Analysis, Yale School of Public Health, New Haven, CT, USA.
7
Department of Ecology and Evolution, Yale University, New Haven, CT, USA.
8
The Santa Fe Institute, Santa Fe, NM, USA.
Received: 1 February 2017 Accepted: 11 April 2017
References
1. Gulland A. Zika virus is a global public health emergency, declares WHO.
BMJ Publishing Group Ltd. 2016;352:i657. doi:10.1136/bmj.i657.
2. Zhang Q, Sun K, Chinazzi M, Pastore-Piontti A, Dean NE, Rojas DP, et al.
Projected spread of Zika virus in the Americas. bioRxiv. 2016;
3. Estimated range of Aedes aegypti and Aedes albopictus in the United States,
2016. In: Atlanta: Centers for Disease Control and Prevention [Internet].
[cited 1 Jan 2016]. Available: http://www.cdc.gov/zika/vector/range.html
4. Florida Department of Health. Department of Health Daily Zika Update
[Internet]. Available: http://www.floridahealth.gov/newsroom/2017/01/
012717-zika-update.html
5. Texas Department of State Health and Human Services. Zika in Texas
[Internet]. 2016. Available: http://www.texaszika.org/.
6. Texas Department of State Health and Human Services. Arbovirus activity in
Texas 2013 surveillance Report. 2014.
7. Texas Department of State Health and Human Services. Arbovirus activity in
Texas 2014 surveillance Report. 2014.
8. Lessler J, Ott CT, Carcelen AC, Konikoff JM, Williamson J, Bi Q, et al. Times to
key events in Zika virus infection and implications for blood donation: a
systematic review. Bull World Health Organ. 2016;94:841–9. doi:10.2471/BLT.
16.174540.
9. Lloyd-Smith JO, Funk S, McLean AR, Riley S, Wood JLN. Nine challenges in
modelling the emergence of novel pathogens. Epidemics Elsevier BV. 2014;
10:35–9. doi:10.1016/j.epidem.2014.09.002.
10. Duffy MR, Chen T-H, Hancock WT, Powers AM, Kool JL, Lanciotti RS, et al.
Zika virus outbreak on Yap Island, Federated States of Micronesia. N Engl J
Med. 2009;360:2536–43. doi:10.1056/NEJMoa0805715.
11. Alfaro-Murillo JA, Parpia AS, Fitzpatrick MC, Tamagnan JA, Medlock J,
Ndeffo-Mbah ML, et al. A cost-effectiveness tool for informing policies on Zika
virus control. PLOS Negl Public Library of Science. 2016;10:e0004743. Available:
http://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0004743
12. PAHO and WHO. Zika-Epidemiological Report - Anguilla; 2016. p. 1–2.
13. Interim CDC recommendations for Zika vector control in the continental
United States. In: Centers for Disease Control and Prevention [Internet]. 2016
[cited 5 Jan 2016]. Available: http://www.cdc.gov/zika/public-health-partners/
vector-control-us.html
14. Office of Travel & Tourism Industries. US Monthly Arrivals Trend Line
[Internet]. 2014 [cited 25 May 2016]. Available: http://travel.trade.gov/view/
m-2014-I-001/index.html
15. Jaynes ET. Information theory and statistical mechanics. Phys Rev American
Physical Society. 1957;106:620–30. doi:10.1103/PhysRev.106.620.
16. Musso D, Cao-Lormeau VM, Gubler DJ. Zika virus: following the path of dengue
and chikungunya? Lancet (London, England). Elsevier; 2015;386: 243–244.
Available: http://www.thelancet.com/article/S0140673615612739/fulltext
17. PAHO and WHO. Zika-Epidemiological Report-Mexico. 2016.
18. Wolsey LA. Integer programming, vol. 42. New York: Wiley New York; 1998.
19. Merow C, Smith MJ, Silander JA. A practical guide to MaxEnt for modeling
species’distributions: what it does, and why inputs and settings matter.
Ecography (cop). Blackwell Publishing Ltd. 2013;36:1058–69. doi:10.1111/j.
1600-0587.2013.07872.x.
20. Alex Perkins T, Siraj AS, Ruktanonchai CW, Kraemer MUG, Tatem AJ, Mlakar J,
et al. Model-based projections of Zika virus infections in childbearing
women in the Americas. Nat Microbiol Nature Publishing Group. 2016;1:
16126. doi:10.1038/nmicrobiol.2016.126.
21. Brady OJ, Johansson MA, Guerra CA, Bhatt S, Golding N, Pigott DM, et al.
Modelling adult Aedes aegypti and Aedes albopictus survival at different
temperatures in laboratory and field settings. Parasit Vectors. 2013;6:351.
doi:10.1186/1756-3305-6-351.
22. Muir LE, Kay BH. Aedes aegypti Survival and dispersal estimated by mark-
release-recapture in northern Australia. AmJTrop Med Hyg. 1998;58:277–82.
Available: http://www.ncbi.nlm.nih.gov/pubmed/9546403
23. Nishiura H, Halstead SB. Natural history of dengue virus (DENV)-1 and DENV-4
infections: reanalysis of classic studies. J Infect Dis Oxford University Press. 2007;
195:1007–13. Available: http://jid.oxfordjournals.org/content/195/7/1007.full
24. Chan M, Johansson MA. The incubation periods of dengue viruses. Vasilakis
N, editor. PLoS One Public Library of Science. 2012;7:e50972. doi:10.1371/
journal.pone.0050972.
25. Black WC, Bennett KE, Gorrochótegui-Escalante N, Barillas-Mury CV,
Fernández-Salas I, de Lourdes MM, et al. Flavivirus susceptibility in Aedes
aegypti. Arch Med Res. 2002;33:379–88. doi:10.1016/S0188-4409(02)00373-9.
26. Cauchemez S, Epperson S, Biggerstaff M, Swerdlow D, Finelli L, Ferguson
NM. Using routine surveillance data to estimate the epidemic potential of
emerging Zoonoses: application to the emergence of US swine origin
influenza a H3N2v virus. Peiris JSM, editor. PLoS Med. 2013;10:e1001399.
doi:10.1371/journal.pmed.1001399.
27. Blumberg S, Lloyd-Smith JO. Comparing methods for estimating R0 from the
size distribution of subcritical transmission chains. Epidemics. 2013;5:131–45.
doi:10.1016/j.epidem.2013.05.002.
28. Rojas DP, Dean NE, Yang Y, Kenah E, Quintero J, Tomasi S, et al. The
epidemiology and transmissibility of Zika virus in Girardot and San Andres
island, Colombia, September 2015 to January 2016. Eur Secur. 2016;21:30283.
doi:10.2807/1560-7917.ES.2016.21.28.30283.
29. Kucharski AJ, Funk S, Eggo RM, Mallet H, Edmunds WJ, Nilles EJ.
Transmission dynamics of Zika virus in island populations : a modelling
analysis of the 2013–14 French Polynesia outbreak. bioRxiv. 2016; 1–15.
30. Texas Department of State Health and Human Services. Texas Announces
Local Zika Virus Case in Rio Grande Valley [Internet]. Available: http://www.
dshs.texas.gov/news/releases/2016/20161128.aspx
31. Dechant E, Rigau-Perez J. Hospitalizations for suspected dengue in Puerto
Rico, 1991-1995: estimation by capture-recapture methods. The Puerto Rico
Association of Epidemiologists. AmJTrop Med Hyg. 1999;61:574–8. Available:
http://www.ajtmh.org/content/61/4/574.short
32. Sidwa TJ. Mosquito surveillance/control in Texas. 2016.
33. Scarpino SV, Iamarino A, Wells C, Yamin D, Ndeffo-Mbah M, Wenzel NS, et
al. Epidemiological and viral genomic sequence analysis of the 2014 ebola
outbreak reveals clustered transmission. Clin Infect Dis Oxford University
Press. 2015;60:1079–82. Available: http://cid.oxfordjournals.org/content/early/
2015/01/05/cid.ciu1131.full
• We accept pre-submission inquiries
• Our selector tool helps you to find the most relevant journal
• We provide round the clock customer support
• Convenient online submission
• Thorough peer review
• Inclusion in PubMed and all major indexing services
• Maximum visibility for your research
Submit your manuscript at
www.biomedcentral.com/submit
Submit your next manuscript to BioMed Central
and we will help you at every step:
Castro et al. BMC Infectious Diseases (2017) 17:284 Page 9 of 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com