Epidemics 17 (2016) 50–55

Contents lists available at ScienceDirect

Epidemics
journal homepage: www.elsevier.com/locate/epidemics

Estimate of the reproduction number of the 2015 Zika virus outbreak
in Barranquilla, Colombia, and estimation of the relative role of sexual
transmission
Sherry Towers a,∗ , Fred Brauer b , Carlos Castillo-Chavez a , Andrew K.I. Falconar c ,
Anuj Mubayi a , Claudia M.E. Romero-Vivas c
a

Arizona State University, Tempe, AZ, USA
University of British Columbia, Vancouver, BC, Canada
c
Universidad del Norte, Barranquilla, Atlantico, Colombia
b

a r t i c l e

i n f o

Article history:
Received 5 June 2016
Received in revised form 11 October 2016
Accepted 12 October 2016
Available online 17 October 2016
Keywords:
Zika virus
Basic reproduction number
Mathematical model
Sexually transmitted infections
Arboviruses

a b s t r a c t
Background: In 2015, the Zika arbovirus (ZIKV) began circulating in the Americas, rapidly expanding
its global geographic range in explosive outbreaks. Unusual among mosquito-borne diseases, ZIKV has
been shown to also be sexually transmitted, although sustained autochthonous transmission due to
sexual transmission alone has not been observed, indicating the reproduction number (R0) for sexual
transmission alone is less than 1. Critical to the assessment of outbreak risk, estimation of the potential
attack rates, and assessment of control measures, are estimates of the basic reproduction number, R0.
Methods: We estimated the R0 of the 2015 ZIKV outbreak in Barranquilla, Colombia, through an analysis
of the exponential rise in clinically identified ZIKV cases (n = 359 to the end of November, 2015).
Findings: The rate of exponential rise in cases was ␳ = 0.076 days−1 , with 95% CI [0.066,0.087] days−1 .
We used a vector-borne disease model with additional direct transmission to estimate the R0; assuming
the R0 of sexual transmission alone is less than 1, we estimated the total R0 = 3.8 [2.4,5.6], and that the
fraction of cases due to sexual transmission was 0.23 [0.01,0.47] with 95% confidence.
Interpretation: This is among the first estimates of R0 for a ZIKV outbreak in the Americas, and also among
the first quantifications of the relative impact of sexual transmission.
© 2016 The Authors. 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/).

1. Introduction
Zika virus (ZIKV), a mosquito-borne arbovirus, was first identified in Uganda in 1947. Similar to the dengue and chikungunya
viruses, ZIKV is primarily spread by the tropical and sub-tropical
domestic mosquito, Aedes aegypti (World Health, 2016). Recent
outbreaks of ZIKV disease have occurred in French Polynesia in
2013-14 (Kucharski, 2016), and Yap Island in the Pacific in 2007
(Duffy et al., 2009), and since 2015, ZIKV has rapidly spread
through many countries in South America, Central America, and
the Caribbean where the Aedes aegypti vector species is endemic
(Fauci and Morens, 2016).

∗ Corresponding author at: Simon A. Levin Mathematical, Computational and
Modeling Sciences Center, Arizona State University, PO Box 873901, Tempe, AZ,
85287-3901, USA.
E-mail addresses: smtowers@asu.edu, sherrytowers@hotmail.com (S. Towers).

ZIKV disease is usually asymptomatic, and is typically mild even
upon clinical presentation, with symptoms similar to that of dengue
virus (DENV) and chikungunya virus (CHIKV) infection (World
Health, 2016; Fauci and Morens, 2016; Hayes, 2009). However,
particular sets of clinical diagnostic criteria have been successfully employed, for example in Brazil (Brasil, 2016) and Colombia
(Tolosa-Perez, 2016), to successfully differentiate patients with
ZIKV infections from DENV and CHIKV infections (Brasil, 2016).
ZIKV disease has been linked to an apparent increased risk of the
neurological disorder Guillain-Barré syndrome, and also to neonate
microcephaly (Fauci and Morens, 2016). The latter is of particular
concern, because pregnant women may not know they have been
infected, and the damage to their unborn infants may result in subsequent lifelong disabilities. In addition, there is evidence that there
is also a direct human sexual transmission component to the disease, although it is unclear how significant a factor this aspect is in
overall transmission in Aedes aegypti infested areas (World Health,
2015). Quantification of the role of ZIKV sexual transmission is cru-

http://dx.doi.org/10.1016/j.epidem.2016.10.003
1755-4365/© 2016 The Authors. 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/).

S. Towers et al. / Epidemics 17 (2016) 50–55

cial to the assessment of the relative efficacy of avoidance of sexual
contact in case reduction.
There is currently no vaccine or specific treatment for ZIKV infection (World Health, 2015), leaving control of the vector populations,
use of mosquito repellents, and avoidance of sexual contact as the
only means to control the spread of the disease.
Critical to the assessment of outbreak risk, and to the design,
development, and evaluation of control strategies, are mathematical models that simulate the underlying dynamics of the
transmission of a disease within a population (Hethcote, 2000). A
particularly important quantity in infectious disease epidemiology
that can be estimated with such models is the basic reproduction number, R0, which is the average number of secondary cases
produced in a completely naïve population by the introduction
of a single infectious individual (Hethcote, 2000; Heffernan et al.,
2005). Very few estimates of the R0 of ZIKV disease have been
published in the literature at the time of this writing (September,
2016), likely largely due to the surveillance challenges posed by
this oftenasymptomatic disease in countries with limited human
and capital resources for surveillance.
Here we examine the outbreak of ZIKV that began in 2015 in Barranquilla, Colombia. Barranquilla is the major Colombian port city
on the Caribbean coast, with a population of 1.4 million (RomeroVivas et al., 2013). It is a center of development in Colombia, and
in recent years has experienced a rapid expansion of urbanization (Romero-Vivas et al., 2013). It has a tropical savannah climate,
with an average daytime temperature of 32 ◦ C year round, with
day/night average of 28.4 ◦ C. Wet seasons occur from April to June
and August to November, with a yearly average precipitation of
just over 800 mm, providing an ideal habitat for the Aedes aegypti
mosquito (Romero-Vivas et al., 2006).
The public health department of the city has had a long-standing
program of surveillance of arboviral diseases in both the human
and vector populations; the dengue viruses have been endemic in
the city for many years, with three serotypes co-circulating prior
to 2007 (DENV-1, -2 and -4), and all four DENV serotypes circulating since then (Falconar et al., 2006; Falconar and Romero-Vivas,
2012). Since 2014, chikungunya virus also emerged in a major outbreak in Barranquilla, and affected a large fraction of the population
(Cardona-Ospina et al., 2015).
In this work we used syndromic surveillance data, and examined the clinically identified cases of ZIKV in Barranquilla from the
beginning of October to the end of December, 2015. We employed
a mathematical model for vector-borne disease transmission with
additional direct sexual transmission to assess the reproduction
number, R0, for ZIKV disease spread, by fitting the parameters of
the model to the initial exponential rise in cases. Given that sexual transmission alone has not resulted in sustained autochthonous
transmission in areas free of the Aedes mosquito vector, we assume
that the R0 corresponding to sexual transmission alone is less than
1. Based on this information, we used our model to assess the fraction of cases in Barranquilla generated by sexual transmission.

2. Methods and materials

51

Table 1
Number of cases of Zika virus disease in Barranquilla, Colombia, with date of identification between October 1, 2015 and December 31, 2015.
Total

n = 1470

Male
Non-pregnant female
Pregnant female
Age 0–17
Age 18–44
Age 45–64
Age 65 and over

610 (41%)
818 (56%)
42 (3%)
232 (16%)
901 (61%)
289 (20%)
48 (3%)

dromic surveillance guidelines (Tolosa-Perez, 2016), as were also
successfully employed in Brazil (Brasil, 2016), and other countries
(see, for instance, (Tognarelli et al., 2015)).
A total of n = 1470 cases were identified between the beginning
of October and the end of December 2015. The gender and age
demographics of those cases are shown in Table 1. The dates of initial symptoms were available for n = 1141 (78%) of these cases. The
incidence data, aggregated by day of initial symptoms, are shown
in Fig. 1.
Of the n = 1470 cases identified in Barranquilla up to the end of
December 2015, 3% occurred in pregnant women. Because concern
over fetal birth defects may have prompted women to seek testing
for ZIKV when they otherwise would not, there was the potential
for surveillance bias. In the following, we thus determine the exponential rise in cases including pregnant women, and crosscheck the
analysis by excluding pregnant women.
2.2. Statistical methods for fitting the exponential rise in
incidence
In the early stages of an infectious disease outbreak, the number
of incident cases grows exponentially in time as the effect of the
increasing incidence on the depletion of the susceptible population
remains small (Wallinga and Lipsitch, 2007).
We employed maximum likelihood methods to fit an exponential curve to the initial rise in daily ZIKV incidence data, by
date of initial symptoms, to determine the initial exponential
growth rate, ␳, using a Negative Binomial likelihood to account
for over-dispersion in the data (Chowell et al., 2012; Lloyd-Smith,
2007). Using the methods described in Chowell et al. (2012), we
determined that the exponential rise in cases occurred up to
approximately the end of November, 2015 (n = 359). Surveillance
during that period by the SIVIGILA system was constant, and no
unusual vector control measures were implemented during that
time.
We crosschecked the robustness of the fit assumptions by fitting
to the initial rise only during the month of October, and again for
the month of November. If the rate of exponential rise was constant
(as it should be when surveillance is constant and no unusual vector
control measures are taken in the initial phase of the outbreak), the
two fits should yield statistically consistent results (Towers et al.,
2014).

2.1. Data

2.3. Estimation of the reproduction number

The data used consisted of the daily incidence of ZIKV disease
cases identified during 2015 in Barranquilla, Colombia, by the syndromic surveillance system of the Colombian National System for
the Public Health (Sistema de Vigilancia en Salud Publica: SIVIGILA) (Ministerio de la, 2016). Human ZIKV infections were clinically
identified upon presentation at the multiple primary public health
clinics within the city as distinct from those caused by DENV or
CHIKV according to the National Health Institute national syn-

ZIKV is a mosquito-borne disease for which a susceptible
mosquito vector species bites an infectious human, whereupon the
virus replicates in the mosquito’s mid gut, and then its salivary
gland cells (Chouin-Carneiro, 2016). After several days (known as
the “extrinsic incubation period”) (Majumder et al., 2016), ZIKV can
be found in the mosquito’s saliva, which then can be transmitted
to other humans that the infected mosquito bites (World Health,
2016). A human, once bitten by an infected mosquito, incubates

52

S. Towers et al. / Epidemics 17 (2016) 50–55

Fig. 1. The time series of clinically identified Zika virus disease cases in Barranquilla, Colombia, with the best-fit exponential curves to the initial rise overlaid.

the virus for several days, whereupon they become infectious for a
period of several more days (Majumder et al., 2016).
In addition to these dynamics, there is evidence that ZIKV can
also be sexually transmitted (Musso et al., 2015; Davidson, 2016;
Dyer, 2016).
In this study, we employed a compartmental mathematical
model to simulate these dynamics (Brauer et al., 2016). The model
included compartments corresponding to Susceptible, Exposed,
Infected, and Recovered humans, and Susceptible, Exposed, and
Infected mosquitoes (and is thus known as an SEIR/SEI model).
In a population of Nh humans and Nv adult female mosquitoes,
the susceptible adult female mosquitoes, Sv, upon biting an infectious human, incubate the virus for an average period, 1/␩, and then
move to the infectious compartment, lv. They die after an average
of 1/␮ days. The transmission rate from humans to mosquitoes is
ˇv .
The susceptible humans, Sh, can be infected by being bitten
by infectious mosquitoes or through direct contact with another
infected human. The human then incubates the virus for an average
period, 1/␬ days, before becoming infectious, Ih. After an average of 1/␥ days, the human then moves to the recovered and
immune compartment, Rh. The transmission rate from mosquitoes
to humans is ˇ, and the direct transmission rate between humans
is ␣.
A similar model (without sexual transmission) was previously
employed to estimate the reproduction number of DENV and CHIKV
fever outbreaks (Chowell et al., 2007, 2008; Yakob and Clements,
2013), and the reproduction number of the 2013-14 outbreak of
ZIKV in French Polynesia in 2007 (Kucharski, 2016).
The compartmental flow diagram for this model is shown in
Fig. 2, and the system of ordinary differential equations incorporat-

Fig. 2. Compartmental flow diagram of the vector-borne and sexual transmission
model described in Eq. (1). Parameters are as described in Table 2.

ing the dynamics of the model were described previously in Brauer
et al. (2016):
Sh ’ = −ˇSh Iv /Nv − ␣Sh Ih /Nh
’

Eh = +ˇSh Iv /Nv + ˛Sh Ih /Nh − Eh
= Eh − Ih
= Ih
Sv = −ˇv Sv Ih /Nh − Nv − Sv
Ih’
Rh’
’

(1)

’

Ev = +ˇv Sv Ih /Nh −  +  Ev
Iv’ = Ev − Iv .
The reproduction number of ZIKV due to vector-borne transmission alone is
R0vector = ˇˇv


,
 ( + )

(2)

S. Towers et al. / Epidemics 17 (2016) 50–55

and the reproduction number of ZIKV due to sexual (or other
direct) transmission alone is R0sex = ˛/. The R0 of the model is
R0 = R0vector + R0sex (Brauer et al., 2016). Here we interpret the RHS
of Eq. (2) as the reproduction number for the vector-borne component of transmission. However, it has been pointed out in the
literature that the RHS of Eq. (2) can equally reasonably be interpreted as R02 , rather than R0 (Brauer et al., 2016) (albeit for the
purposes of this paper, it does not matter which form we use). The
form in Eq. (2) is referred to as the “type reproduction number”
and is the total number of secondary infections in humans originating from a human infection. In practice, we note that several past
SEIR/SEI model estimates of the reproduction number for DENV,
CHIKV, and ZIKV outbreaks have also used the formulation as we
have it here (Chowell et al., 2007, 2008; Yakob and Clements, 2013);
for consistency of comparison between our result and those estimates, we thus used the formulation as we have presented. But
it must generally be kept in mind when comparing our results to
those of other analyses that the explicit form of the reproduction
number expression must be noted.
The expression relating the total reproduction number to the
rate of exponential rise, ␳, and the R0sex is, based on the results
presented in Reference (Brauer et al., 2016),

R0 = R0sex +





( + ) ( + )
( + ) ( +  + )
− R0sex
.

 ( + )

(3)

Thus, given the rate of exponential rise in cases, an estimate
of the R0 of sexual transmission, and estimates of the latent and
infectious periods and the mosquito lifespan, we can estimate the
reproduction number. Currently, there is no estimate of R0sex , however we assume from the lack of sustained ZIKV transmission in
areas free of the Aedes vector that R0sex < 1.
We note here that ZIKV has a high rate of asymptomatic infection, thus the observed cases represent a small fraction of the true
number (World Health, 2016; Fauci and Morens, 2016). However,
this scale factor does not change the rate of exponential rise in
cases; thus estimation of the rate of exponential rise with only
partial observation of cases is a robust means to estimate the reproduction number.
To estimate the probability distribution for R0, given probability
distributions for R0sex , ␳, ␬, ␥, ␩, and ␮, we performed ten thousand
Monte Carlo iterations, whereby we randomly sampled values of ␳
and the model parameters from their probability distributions, and
calculated the resulting estimate of R0 at each iteration using Eq.
(3), and also the fraction of cases due to sexual transmission estimated from the numerical solution to Eq. (1), both for the entire
simulated epidemic curve, and also the exponential rise region,
identified using the methods of Reference (Chowell et al., 2012).
The serial interval of ZIKV is the sum of the incubation periods plus
half the infectious periods, T = 1/␬ + 1/␩ + 0.5/␮ + 0.5/␥, as described
in Reference (Svensson, 2007); we thus used the constraint that
the serial interval derived from the sampled parameters had to be
within the observed serial interval of 10 to 23 days (Majumder et al.,
2016). The resulting distribution of the estimates of R0 formed the
estimate of the probability density for that quantity (Gardner and
O’Neill, 1983).
We currently know that R0sex < 1, but do not know its value, thus
the probability distribution used for R0sex was Uniform(0,1). The
probability distributions for ␬, ␥, ␩, and ␮ were derived from the
ranges of the literature estimates for these quantities, assuming a
Uniform probability distribution over the range. The estimates used
to assess the range of the Uniform probability distributions for these
parameters are shown in Table 2. The probability distribution for
the rate of exponential rise, ␳, was derived from the likelihood fit to
the data, and was Normally distributed about the best-fit estimate.

53

3. Results
3.1. Estimation of the rate of initial exponential rise in cases
The results of the fit of an exponential rise to the daily incidence data of clinically identified ZIKV cases in Barranquilla from
the beginning of October up to the end of November (n = 359) are
shown overlaid on the data in Fig. 1. The estimated rate of exponential rise was ␳ = 0.076 days−1 , with 95% confidence interval [0.066,
0.087] days−1 .
We crosschecked our assumption of a constant rate of exponential rise in the initial data by fitting only up to the end of October
(n = 32), yielding ␳ = 0.111 [0.060, 0.167] days−1 . We additionally fit
to the data from the beginning to the end of November (n = 327),
yielding ␳ = 0.071 [0.055, 0.089] days−1 . These two rates of rise were
statistically consistent (␹2 ␯ = 1 test p = 0.15), thus the exponential
rate of rise in ZIKV cases appeared to be consistent over the entire
period from the beginning of October to the end of November. These
two fits are also shown overlaid on the incidence data in Fig. 1.
Excluding pregnant women from the fit only excluded three
cases up to the end of November, and resulted in negligible changes
to the estimated exponential rise.
3.2. Estimation of the reproduction number
With the use of Eq. (1), we obtained the average value of the estimated reproduction number as R0 = 3.8 with [2.4,5.6] 95% CI, and
the fraction of cases due to sexual transmission was 0.23 [0.01,0.47],
with 95% confidence. The latter quantity was estimated from the
entire simulated epidemic curve. In the exponential rise region
only, the fraction of cases was 0.27 [0.01,0.56], with 95% confidence
4. Discussion
We have estimated the R0 of a ZIKV outbreak occurring in
an area typical of the regions in the Americas that are currently
affected by the disease.
Despite the pressing need to estimate the R0 for a newly emerging pandemic disease, there are currently only two other estimates
of the R0 of a ZIKV disease outbreak of which the authors are
aware; the R0 of the 2013-14 ZIKV outbreak in French Polynesia
was recently estimated to be between 2.6 to 4.8 (Kucharski, 2016),
in statistical agreement with the estimate of the R0 obtained by
this analysis, R0 = 3.8 [2.4, 5.6]. The R0 of ZIKV based on data from
Colombia, Brazil, and El Salvador at the beginning of the 2015/2016
outbreak was estimated to be R0 = 2.055 [0.523,6.300] with an estimated percentage of cases due to sexual transmission being 3.044%
[0.123%,45.73%] (Gao et al., 2016). This estimated percentage of
cases due to sexual transmission is in statistical agreement with our
estimate, that the percentage was 23 [1,47]% with 95% confidence.
Note that the vector-borne R0 used in Reference (Gao et al., 2016)
is the square root of the type-reproduction number we use, thus
the results in Reference (Gao et al., 2016) for R0 must be approximately squared for comparison to those of this analysis. However,
due to the very large confidence interval on R0 estimated by Reference (Gao et al., 2016), both the squared and un-squared values
are in statistical agreement with our estimate. We note here, as an
aside, that the confidence interval for R0 estimated by Reference
(Gao et al., 2016) paradoxically includes values less than one, even
though the estimates were obtained by fitting to data from three
large epidemics, thus values of R0 less than 1 should have been
impossible.
SEIR/SEI model estimates for the R0 of DENV outbreaks generally
fall between approximately 1.5 to 3 (see, for instance, References
(Chowell et al., 2007, 2008)), and around 4 for CHIKV outbreaks

54

S. Towers et al. / Epidemics 17 (2016) 50–55

Table 2
Model parameters used in this analysis. The incubation and infectious periods, and the mosquito average lifespan, were determined from the cited references, and the
references therein.
Parameter

Definition

Estimate (days)

Reference

1/␬

Intrinsic (human)
latent period

3–12
2–7
2–6

(Majumder et al., 2016)
(World Health, 2016) (Kucharski, 2016)

Assume Uniform (2,12)
1/␥

Human infectious
Period
Assume Uniform (3,7)

3–5
3–7

(Majumder et al., 2016)
(Kucharski, 2016)

1/␩

Extrinsic (mosquito)
infectious period

4–6
10–15
8–13

(Majumder et al., 2016)
(Hayes, 2009) (Kucharski, 2016)

6–15
10–20
6–10

(Chowell et al., 2008, 2007)
(Yakob and Clements, 2013) (Kucharski, 2016)

10–23

(Majumder et al., 2016)

Assume Uniform (4,15)
1/␮

Average mosquito
lifespan
Assume Uniform (6,20)

1/␩ + 1/␬ + ½ ␥ + ½ ␮

Serial Interval

(Yakob and Clements, 2013; Svensson, 2007). The estimated R0 for
the ZIKV outbreak in Barranquilla is similar to that of CHIKV outbreaks, but is larger than the estimates for DENV outbreaks, despite
the fact that the vector species is the same for the diseases. This
could be due to a variety of reasons, including a sexual transmission component to ZIKV (Fauci and Morens, 2016; Musso et al.,
2015), and/or partial prior immunity to DENV in hyperendemic
areas, and/or shorter incubation periods for ZIKV in the vectors,
and/or a longer infectious period in humans, or even increased
susceptibility to infection.
It is important to note here that our estimate of the ZIKV R0 is
model dependent, as indeed are all such estimates in the literature.
However, for ease of comparison of results we have used a similar model to that employed to analyze the French Polynesia ZIKV
outbreak (Kucharski, 2016), DENV outbreaks (Chowell et al., 2007,
2008), and a CHIKV outbreak (Yakob and Clements, 2013). While
we have not explored other model formulations in this analysis, it
would be easy to use our estimate of the exponential rise with other
compartmental model formulations, and/or other model parameter
values, in order to extract alternative estimates of R0.
Our model assumed homogeneity in contacts between members of the population, similar to other studies (Gao et al., 2016).
In reality, contacts between members in the population, especially
sexual contacts, are generally highly networked and gender dependent. However, early in an epidemic during the exponential growth
phase, the rate of exponential growth is driven largely by the R0 and
serial interval of the disease (Barthélemy et al., 2005; Brauer, 2012),
and it is only later in the epidemic that model mixing assumptions affect forecasts of the outbreak. Thus, despite the simplified
assumptions of mixing in our model we can estimate the R0, but
the model should not be used to estimate the final size of the epidemic, or to quantify the efficacy of control strategies, including
those aimed at sexual transmission.
There is currently much uncertainty related to the epidemiological parameters of ZIKV infection, such as the human and mosquito
latent periods, and R0 of sexual transmission, which are almost
wholly responsible for the somewhat broad confidence interval
on our estimate of R0. Further study will help to constrain these
parameters, and allow for more precise estimates of the R0. Again,
because we have provided our estimate of the exponential rise in
cases for this outbreak, it would be easy in the future to re-calculate
the R0 once these parameters are better known.
Our study is based on clinical syndromic surveillance data, similar to many other studies of arboviral outbreaks (for instance,
References (Cardona-Ospina et al., 2015; Chowell et al., 2007,

2008)), due to limited laboratory testing resources in developing
countries where many of these outbreaks occur. ZIKV has similar
symptoms to some of the symptoms seen in mild cases of DENV
(Brasil, 2016), and further study is needed in the future to determine
how potential misdiagnosis may affect apparent temporal dynamics in places where other arboviral diseases may be co-circulating
(Fauci and Morens, 2016). However, it must be pointed out that initial studies indicate very low frequency of contamination by other
arboviral infections in cases identified by ZIKV clinical surveillance in areas where DENV and CHIKV co-circulate (Brasil, 2016;
Tognarelli et al., 2015).
Changes in human behavioral responses, including healthseeking behaviors, are always a potential confounder in quantitative studies of outbreaks of high profile newly emerging diseases
when fear may be incited in the population (see, for example, References (Fenichel et al., 2011; Lau et al., 2005; Plucinski et al.,
2015)). However, the data we consider in this analysis in October
and November 2015 are early in the outbreak, largely before significant international attention began to be paid to ZIKV and the
birth defects and other disorders it can cause; we note here that
the first epidemiological alert by the WHO and the Pan American
Health Organisation detailing a possible link between ZIKV infection and microcephaly was on November 17, 2015.1 After the period
considered during this study, there was a sharp increase in the
number of cases identified in pregnant women, potentially indicating increasing concern in this group as time went on; during
the month of December, 39 of the 782 (5%) identified cases with
dates of initial symptoms were pregnant females, which is not statistically consistent with the 3 of the 359 (1%) cases considered
during October and November (␹2 ␯ = 1 p < 0.001). In addition, the
fraction of cases involving pregnant women during October and
November 2015 was statistically consistent with the crude birth
rate in Colombia (16/1000),2 indicating no apparent heightened
concern in that group during that period.
Sexual transmission has been noted to play a role in ZIKV transmission (World Health, 2015), but the relative contribution of
sexual transmission cannot be determined based on exponential
rise in incidence alone unless further information is known. In this
case, we use the fact that sustained transmission in areas without
Aedes mosquitoes has not been observed, thus the R0 of sexual or

1
http://www.who.int/bulletin/online first/16-171082/en/,
2016.
2
www.data.worldbank.org, accessed September, 2016.

accessed

August,

S. Towers et al. / Epidemics 17 (2016) 50–55

other direct transmission must be less than 1. Our estimated percentage of cases due to sexual transmission when Aedes mosquitoes
are also present may be as high as 47%, with 95% confidence. As
such, safe sexual practices may significantly reduce incidence.
5. Conclusions
ZIKV has spread explosively in the Americas in recent months.
Here we employed an SEIR/SEI mathematical model of the spread
of vector-borne disease with additional sexual transmission to estimate the basic reproduction number, R0, of the ZIKV outbreak that
began in October 2015, in Barranquilla, Colombia. We estimated the
R0 to be R0 = 3.8 with 95% CI [2.4,5.6], and a one standard deviation
uncertainty of 0.8. This is among the first estimates of the R0 of a
ZIKV outbreak in the Americas, and is important to assessment of
outbreak risk in new areas. Our modeling analysis estimates that up
to 47% of ZIKV cases in Barranquilla, with 95% confidence, may have
been due to sexual contact alone. Thus safe sexual practices may
significantly reduce incidence during ZIKV outbreaks in tropical and
semi-tropical areas.
Acknowledgments
The authors are very grateful to Dr. Elsa Bravo de Plata and Dr.
Pedro Arango of the Secretaria Distrital de Salud de Barranquilla,
Colombia, for their advice and assistance related to these studies.
The authors are grateful for the following sources of funding that
supported these studies:
FB: Natural Sciences and Engineering Research Council of
Canada, Grant OGPIN 203901-99;
AKIF, CMERV: Proyecto de Regalias, Convenio0103-2015000039, Actividad MGA [955869];
CCC, ST, AM: National Science Foundation (DMS-1263374,
DUE-1101782), National Security Agency (H98230-14-1-0157), the
Office of the President of ASU, and the Office of the Provost of ASU.
References
Barthélemy, M., et al., 2005. Dynamical patterns of epidemic outbreaks in complex
heterogeneous networks. J. Theor. Biol. 235 (2), 275–288.
Brasil, P., et al., 2016. Zika virus outbreak in Rio de Janeiro, Brazil: clinical
characterization, epidemiological and virological aspects. PLoS Negl. Trop. Dis.
10 (4), e0004636.
Brauer, F., 2012. Heterogeneous mixing in epidemic models. Can. Appl. Math. Q. 20
(1), 1–13.
Brauer, F., et al., 2016. Some models for epidemics of vector-transmitted diseases
Infectious Disease Modelling, 2016, Aug 30.
Cardona-Ospina, J.A., et al., 2015. Estimating the burden of disease and the
economic cost attributable to chikungunya, Colombia, 2014. Trans. R. Soc.
Trop. Med. Hyg. 109 (12), 793–802.
Chouin-Carneiro, T., 2016. Differential susceptibilities of aedes aegypti and aedes
albopictus from the americas to zika virus. PLoS Negl. Trop. Dis. 10 (3),
e0004543.
Chowell, G., et al., 2007. Estimation of the reproduction number of dengue fever
from spatial epidemic data. Math. Biosci. 208 (2), 571–589.
Chowell, G., et al., 2008. Spatial and temporal dynamics of dengue fever in Peru:
1994–2006. Epidemiol. Infect. 136 (12), 1667–1677.
Chowell, G., et al., 2012. The influence of climatic conditions on the transmission
dynamics of the 2009 A/H1N1 influenza pandemic in Chile. BMC Infect. Dis. 12
(1), 298.

55

Davidson, A., 2016. Suspected female-to-male sexual transmission of zika
virus—New York city, 2016. Morbidity and Mortality Weekly Report.
MMWR65.
Duffy, M.R., et al., 2009. Zika virus outbreak on Yap Island: federated states of
Micronesia. New Engl. J. Med. 360 (24), 2536–2543.
Dyer, O., 2016. US probes first apparent non-sexual person-to-person Zika
transmission and first domestic outbreak. Br. Med. J. 35, i4107.
Falconar, A.K., Romero-Vivas, C.M., 2012. Simple prognostic criteria can
definitively identify patients who develop severe versus non-severe dengue
disease, or have other febrile illnesses. J. Clini. Med. Res. 4 (1), 33.
Falconar, A.K., de Plata, E., Romero-Vivas, C.M., 2006. Altered enzyme-linked
immunosorbent assay immunoglobulin M (IgM)/IgG optical density ratios can
correctly classify all primary or secondary dengue virus infections 1 day after
the onset of symptoms: when all of the viruses can be isolated. Clin. Vaccine
Immunol. 13 (9), 1044–1051.
Fauci, A.S., Morens, D.M., 2016. Zika virus in the americas—yet another arbovirus
threat. New Engl. J. Med.
Fenichel, E.P., et al., 2011. Adaptive human behavior in epidemiological models.
Proc. Natl. Acad. Sci. 108 (15), 6306–6311.
Gao, D., et al., 2016. Prevention and control of Zika fever as a mosquito-borne and
sexually transmitted disease. Sci. Rep. 6, 28070.
Gardner, R., O’Neill, R., 1983. Parameter Uncertainty and Model Predictions: a
Review of Monte Carlo Results, in Uncertainty and Forecasting of Water
Quality. Springer, pp. 245–257.
Hayes, E.B., 2009. Zika virus outside africa. Emerg. Infect. Dis. 15 (9), 1347.
Heffernan, J., Smith, R., Wahl, L., 2005. Perspectives on the basic reproductive ratio.
J. R. Soc. Interface 2 (4), 281–293.
Hethcote, H.W., 2000. The mathematics of infectious diseases. SIAM Rev. 42 (4),
599–653.
Kucharski, A.J., et al., 2016. Transmission dynamics of Zika virus in island
populations: a modelling analysis of the 2013–14 French Polynesia outbreak
bioRxiv, p. 038588.
Lau, J.T., et al., 2005. Impacts of SARS on health-seeking behaviors in general
population in Hong Kong. Prev. Med. 41 (2), 454–462.
Lloyd-Smith, J.O., 2007. Maximum likelihood estimation of the negative binomial
dispersion parameter for highly overdispersed data, with applications to
infectious diseases. PLoS One 2 (2), e180.
Majumder, M.S., et al., 2016. Estimating a feasible serial interval range for Zika
fever. Bull. World Health Organ., 1–6, BLT.16.171009.
Ministerio de la, Proteccion Social, Reglamentacion del Sistema de Vigilancia en
Salud Publica en Colombia (accessed February, 2016), 2006.
Musso, D., et al., 2015. Potential sexual transmission of Zika virus. Emerg. Infect.
Dis. 21 (2), 359.
Plucinski, M.M., et al., 2015. Effect of the Ebola-virus-disease epidemic on malaria
case management in Guinea, 2014: a cross-sectional survey of health facilities.
Lancet Infect. Dis. 15 (9), 1017–1023.
Romero-Vivas, C., Arango-Padilla, P., Falconar, A., 2006. Pupal-productivity surveys
to identify the key container habitats of Aedes aegypti (L.) in Barranquilla, the
principal seaport of Colombia. Ann. Trop. Med. Parasitol. 100 (Supplement-1),
87–95.
Romero-Vivas, C.M., et al., 2013. Cross-sectional study of Leptospira
seroprevalence in humans, rats, mice, and dogs in a main tropical sea-port city.
Am. J. Trop. Med. Hyg. 88 (1), 178–183.
Svensson, A., 2007. A note on generation times in epidemic models. Math. Biosci.
208 (1), 300–311.
Tognarelli, J., et al., 2015. A report on the outbreak of zika virus on easter island,
South Pacific, 2014. Arch. Virol., 1–4.
Tolosa-Perez, N., 2016. Protocolo de vigilancia en Salud Pública: Enfermedad por
Virus Zika. Instituto Nacional de Salud/Ministerio de Salud de Colombia.
Towers, S., Patterson-Lomba, O., Castillo-Chavez, C., 2014. Temporal variations in
the effective reproduction number of the 2014 West Africa Ebola outbreak.
PLoS Curr. 6.
Wallinga, J., Lipsitch, M., 2007. How generation intervals shape the relationship
between growth rates and reproductive numbers. Proc. R. Soc. Lond. B: Biol.
Sci. 274 (1609), 599–604.
World Health Organization, 2015. (accessed February, Zika virus.
World Health Organization, 2016. Zika virus factsheet (accessed February 2016).
Yakob, L., Clements, A.C., 2013. A mathematical model of Chikungunya dynamics
and control: the major epidemic on Reunion Island. PLoS One 8 (3), e57448.