Predicting and Evaluating the Epidemic Trend of Ebola Virus Disease in the 2014-2015 Outbreak and the Effects of Intervention Measures
et al. (2016) Predicting and Evaluating the
Epidemic Trend of Ebola Virus Disease in the 2014-
2015 Outbreak and the Effects of Intervention
Measures. PLoS ONE 11(4): e0152438. doi:10.1371/
Predicting and Evaluating the Epidemic Trend of Ebola Virus Disease in the 2014-2015 Outbreak and the Effects of Intervention Measures
Zuiyuan Guo 0 1
Dan Xiao 1
Dongli Li 0 1
Xiuhong Wang 0 1
Yayu Wang 0 1
Tiecheng Yan 0 1
Zhiqi Wang 1
0 Department of Disease Control, Center for Disease Control and Prevention of Shenyang Military Region , Shenyang, Liaoning Province , China , 2 Department of Epidemiology, Fourth Military Medical University , Xi'an, Shaanxi Province , China , 3 College of Municipal and Environmental Engineering, Shenyang Jianzhu University , Shenyang, Liaoning Province , China
1 Editor: Zhen Jin, Shanxi University , CHINA
We constructed dynamic Ebola virus disease (EVD) transmission models to predict epidemic trends and evaluate intervention measure efficacy following the 2014 EVD epidemic in West Africa. We estimated the effective vaccination rate for the population, with basic reproduction number (R0) as the intermediate variable. Periodic EVD fluctuation was analyzed by solving a Jacobian matrix of differential equations based on a SIR (susceptible, infective, and removed) model. A comprehensive compartment model was constructed to fit and predict EVD transmission patterns, and to evaluate the effects of control and prevention measures. Effective EVD vaccination rates were estimated to be 42% (31-50%), 45% (4248%), and 51% (44-56%) among susceptible individuals in Guinea, Liberia and Sierra Leone, respectively. In the absence of control measures, there would be rapid mortality in these three countries, and an EVD epidemic would be likely recur in 2035, and then again 8~9 years later. Oscillation intervals would shorten and outbreak severity would decrease until the periodicity reached ~5.3 years. Measures that reduced the spread of EVD included: early diagnosis, treatment in isolation, isolating/monitoring close contacts, timely corpse removal, post-recovery condom use, and preventing or quarantining imported cases. EVD may re-emerge within two decades without control and prevention measures. Mass vaccination campaigns and control and prevention measures should be instituted to prevent future EVD epidemics.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information files
Competing Interests: The authors have declared
that no competing interests exist.
The first case of Ebola Virus Disease (EVD) was detected in Guinea in December 2013. Within
months, the disease had spread to Liberia, Sierra Leone, Nigeria, and other countries [
World Health Organization (WHO) was notified officially of the rapid evolution of the EVD
outbreak on March 23, 2014 . A total of 26,277 EVD cases and 10,884 EVD-related deaths
were reported from December 2013 to April 29, 2015 [
]. EVD was transmitted by direct and
indirect contact between people. The mean incubation period was 11.4 days, and typical
symptoms included fever, nausea, vomiting, diarrhea, and generalized pain [
]. The case fatality
rate in the 2014 EVD outbreak was approximately 64% among patients who received
treatment, and approximately 71% among all known infected patients [
]. The spread of the EVD
outbreak was overwhelming and posed a serious threat to public health in Africa due to the
high infection rate of EVD, ineffective control measures during the early-onset stage, local
funeral customs, and weak epidemic prevention measures [
Understanding the epidemiological distribution and transmission pattern of EVD during
the African outbreak will contribute to improve scientific understanding and to allow the
development of more effective control and prevention measures. Other scholars have already
analyzed the epidemiological distribution and performance of clinical centers [
3, 5, 13
providing valuable information for better understanding the EVD outbreak. However, there is no
recognized mathematical model that describes the process of EVD transmission, and that can be
used to evaluate the effect of prevention measures. This lack of a model might hinder the
process of controlling the disease.
A dynamic transmission model of an infectious disease can be constructed by using the
patterns of disease occurrence and transmission, and the relationships between the disease and
social factors. Such a model is used to analyze the dynamic characteristics of disease
transmission quantitatively, to indicate transmission patterns, and to predict the epidemiological trends
of the disease. Dynamic transmission models have been constructed successfully for SARS
(severe acute respiratory syndrome), AIDS (acquired immune deficiency syndrome), and
malaria to analyze transmission patterns, predict epidemiological trends, and evaluate the
effect of intervention measures [
]. These models have provided important evidence for
the scientific control and prevention of these diseases. In a previous study, researchers
constructed an SEIR (susceptible, exposed, infective, removed) model to analyze the 1995 EVD
outbreak in the Congo. The constructed model was well fitted and demonstrated to be effective
for prediction [
], indicating that a dynamic transmission model can be applied to the analysis
of EVD outbreaks. However, the transmission patterns of infectious diseases are affected by
many factors and can be variable across different time periods and regions. Therefore, the 1995
outbreak model may not be applicable to the 2014 outbreak, and it is necessary to construct an
updated model based on the 2014–2015 transmission patterns and the factors that might
Here, we constructed dynamic transmission models for EVD based on the relationships
among people in each stage of the disease. The results may aid in predicting epidemiological
trends, evaluating the effectiveness of intervention measures, and providing scientific evidence
for improving the control and prevention of EVD.
Estimation of the effective vaccination rate
The R0 values and corresponding 95% confidence intervals for EVD in Guinea, Liberia, and
Sierra Leone were 1.71 (1.44–2.01), 1.83 (1.72–1.94), and 2.02 (1.79–2.26), respectively [
2 / 17
Fig 1. Effects of EVD on total population and EVD case number in the absence of control and
prevention measures. Changes in the total number of individuals in the population and in the number of
EVD cases in Guinea (A, B), Liberia (C, D), and Sierra Leone (E, F) are shown. The x-axes indicate time in
months, and the y-axes indicate population number or the number of EVD cases. The shaded area indicates
the 95% confidence interval.
Therefore, at least 42% (31–50%), 45% (42–48%), and 51% (44–56%) of individuals in these
respective countries should be immune to EVD.
Epidemic trends of EVD in the absence of control and prevention measures
Based on the SIR model, without any outbreak control or prevention measures, the total
population number would decrease rapidly in a short period of time (Fig 1). The net growth
threshold values for the epidemic in Guinea (0.95), Liberia (1.01), and Sierra Leone (0.90) were all
close to 1.0. Therefore, when the parameters of the model are held constant over a sufficiently
long time period, the populations of the three countries decrease to constant levels, and the
numbers of individuals in the s, i, and r blocks would oscillate with a periodic weakly
3 / 17
Fig 2. Periodical low dampened oscillation of s and i around the dynamic equilibrium point. The proportions of EVD cases (left y-axis) and susceptible
people (right y-axis), compared to the total population, are shown as functions of time in years. In the absence of control and prevention measures, the s and i
values would oscillate periodically around their dynamic equilibrium points. Meanwhile, periods between oscillations would shorten, and amplitudes would
dampened oscillation, and reach a positive dynamic equilibrium. The equilibrium points for s,
i, and r were found to be 58.5%, 0.1%, and 41.1%, respectively, in Guinea, 54.6%, 0.1%, and
45.3%, respectively, in Liberia, and 49.5%, 0.1%, and 50.4%, respectively, in Sierra Leone.
Combining the three countries into a single region, the model indicates that if there were no
EVD outbreak control or prevention measures in place, EVD would first re-emerge in 2035,
and then re-emerge again about 8–9 years later. The oscillation periods would shorten and the
amplitudes (severity of EVD epidemic) would decrease gradually until the oscillation occurred
with a sustained periodicity of approximately 5.3 years around the positive dynamic
equilibrium point (Fig 2).
Predicting the numbers of EVD cases and deaths
The WHO has reported numbers of EVD cases (confirmed and suspected) and deaths due to
]. These data were used to compare the predicted values from the model to the actual
data. The infection rate for patients in an uncontrolled environment (βI), indicates the number
of susceptible individuals to be infected by an EVD patient per unit time following a gamma
distribution pattern; meanwhile, the βD, βP, and βR rates reflect an exponential distribution
pattern. The numbers of EVD cases and EVD-related deaths predicted by the model were similar
to the actual data (Fig 3), indicating that the model simulated the epidemic pattern of EVD
4 / 17
Fig 3. Comparison of the numbers of EVD cases and deaths predicted by the model to those
published by the WHO. The cumulative number of EVD cases and deaths are plotted as a function of time t
in months. The cumulative number of cases predicted curve (blue) includes confirmed and suspected cases.
Each red point indicates the number of cumulative cases (blue curve) or deaths due to EVD (green curve)
published by the WHO.
Modeling the effects of isolating close contacts on the epidemic trend of EVD
Individuals in close contact with EVD patients are a high-risk population and are prone to
infecting other susceptible individuals if they are not isolated and closely monitored. Therefore,
quarantining the close contacts of EVD patients early is critical to control and prevent
transmission of EVD. As shown in Fig 4A, quarantining close contacts of patients decreased the predicted
incidence of EVD dramatically. By May 31st, 2015, the cumulative number of EVD cases predicted
without isolation of close contacts was 37% greater than the predicted number with isolation.
Effects of decontaminating the corpses of dead EVD patients on the epidemic trend of EVD
Ebola virus is stored in human body fluids. Thus, to prevent the spread of disease, it is critical
to take timely measures to decontaminate the corpses of dead EVD patients. Examples of these
measures would include burning or deep burial. In contrast, the West African funeral custom
of touching the corpses is prone to spreading the disease. The number of cumulative EVD
cases could be decreased by shortening the interval between death and corpse decontamination
5 / 17
Fig 4. Effects of control and prevention measures on EVD epidemic trends. (A) Summary of the effect of
isolating EVD patients’ close contacts on the cumulative number of EVD cases. Note the markedly lower
curve with close contact quarantine (red) relative to that without close contact quarantine (blue). (B) Predicted
numbers of cumulative EVD cases when corpses of dead EVD patients were allowed to stay in the
surrounding environment for 3 (blue), 2 (green), 1 (red), or 0.5 days (yellow). (C) Predicted new EVD cases
when the average symptom onset-to-isolation and treatment is 5 days (blue), 4 days (green), and 3 days
(red). (D) Predicted reduction in the number of new EVD cases from August 1st, 2014 to May 31st, 2015 when
recovered patients have condom-protected (red) versus unprotected (green) sex. (E) Summary of the effect
of cases being imported from other countries on the epidemic trend of EVD. Note that introduction of a new
case every other day (0.5 cases/day, green) has a strong effect on amplitude relative to zero case importation
(red). In all panels, the number of predicted EVD cases is plotted as a function of time.
(Fig 4B). Our dynamic model indicates that, reduction of the interval between a person’s death
and clearing of the corpse to less than 3 days could have reduced the cumulative number of
cases by 19% (2 day interval), 34% (1 day interval), or as much as 41% (0.5 day interval)
through May 31st, 2015.
6 / 17
Effects of altering the time from symptom onset to treatment in isolation on the epidemic trend of EVD
The average time from symptom onset to medical treatment in isolation for EVD patients is 5
]. During this time, EVD patients have a high probability of infecting susceptible
individuals. Early detection, diagnosis, and isolation can decrease the total number of EVD cases in
an open environment, and early treatment can increase the EVD survival rate. A shorter
interval between symptom onset and medical treatment in isolation led to fewer EVD cases and an
earlier peak in the epidemic curve (Fig 4C). The dynamic model indicates that: if the average
interval between symptom onset and treatment were shortened from 5 to 4 days (60%
reduction) or 3 days (87% reduction), the number of EVD cases at the peak of the epidemic curve
would be markedly decreased in amplitude, and shifted leftward to 16 or 40 days earlier,
Effects of recovered EVD patients practicing protected sex on the epidemic trend of EVD
Ebola virus is detectable in the semen of recovered individuals and can be transmitted through
sexual contact [
]. Although sexual transmission is not the major route of transmission for
EVD, it can still impact the epidemic trend. If recovered patients practiced protected sex using
a condom, then the number of EVD cases is reduced and the peak of the epidemic curve is
advanced (Fig 4D). The dynamic model indicates that if all recovered individuals had practiced
protected sex from August 1st of 2014 onward, then the number of EVD cases at the peak of
the epidemic curve would have been reduced by 26% and occurred 19 days earlier.
Effects of cases imported from other countries on the epidemic trend of EVD
Finally, we analyzed the influence of cases being imported from other countries on the epidemic
trend of EVD with the dynamic model. If 0.5 cases were being imported from the uncontrolled
outside environment daily, then the number of new EVD cases at the peak of the epidemic
curve would be 1.6 times higher than that in the absence of any imported cases (Fig 4E). This
finding indicates that imported cases can contribute substantially to EVD epidemic trends.
The sensitivity analysis was conducted with 15 parameters and a continuous time-series for
confirmed and suspected cases. We took 1000 samples from a uniform distribution of each
parameter range of the comprehensive compartment model. Partial rank correlation
coefficients (PRCCs) of the parameters near -1 or +1 indicate a strong impact on the output.
Parameters αi and βi had particularly high PRCCs (Fig 5), which means that the output of this model
is highly sensitive to changes in βI.
An accurate analysis of the transmission pattern of the 2014 EVD epidemic provides a basis for
implementing control and prevention measures. In this study, we constructed a dynamic
transmission model to predict the epidemic trend of EVD and evaluated the effects of control and
prevention measures. The key parameter in the model was the infection rate for patients in the
free environment (βI). This parameter was influenced by a patient’s sphere of activity, the
virulence of the virus, and the proportion of susceptible people in the total population. Fluctuations
in βI may cause the results of the model to vary. Thus, it is important to fit an infection rate
7 / 17
Fig 5. Sensitivity analysis for confirmed and suspected cases. (A) PRCCs that stay positive (>0) most of
the time. (B) PRCCs that stay negative (<0) most of the time. (C) PRCCs that tend to stay near 0.
that approximates real values closely. Here, we found that the distribution of βI was
approximated by a gamma distribution, rather than by an exponential distribution as an earlier study
]. This discrepancy may be due to differences in the nature of the EVD epidemics
modeled. In the 2014 epidemic, most of the early EVD cases were located in the sparsely
8 / 17
populated rural areas, and the patients had a limited sphere of activity. When the disease
spread to urban areas, patients came into contact with more susceptible individuals, causing
the βI value to increase. Subsequently, isolation measures for EVD cases and improvements in
public efforts at self-protection reduced the number of close contacts between EVD patients
and the susceptible population, ultimately reducing the βI value.
Vaccination is an important means of preventing infectious disease [
]. Although there are
no licensed EVD vaccines, we can estimate the demand for a vaccine, which may contribute to
the formation of a vaccine and immunization program development. We estimated minimum
effective vaccination rates for each country, and found that EVD would be controlled if greater
than 42%, 45%, and 51% of the total population had immunity against the Ebola virus in
Guinea, Liberia, and Sierra Leone, respectively. These results provide scientific evidence for
vaccine production and application.
Assuming that no control and prevention measures were taken, understanding the effects of
an unchecked EVD outbreak on the total population number would help us better appreciate
how harmful EVD has the potential to be. Using an SIR model, we found that the population
would be dramatically decreased in a short time in the absence of control and prevention
measures. These results support an overwhelming risk from an EVD outbreak. The SIR model also
indicated that the s, i, and r populations fluctuate periodically around their dynamic
equilibrium points. These results support the idea EVD epidemics may break out periodically, though
the disease will reach a dynamic equilibrium point with a shorter period between epidemics of
progressively weaker intensity. Periodic fluctuations of EVD epidemics are similar to those of
other diseases, such as the measles and chickenpox [
]. Therefore, we should be vigilant
for the reemergence of EVD.
The dynamic model constructed for the 2014–2015 EVD outbreak indicated that early
detection, diagnosis, and isolation are critical for controlling EVD outbreaks. From March
2014 through January 2015, the mean time interval between symptom onset and sample
collection from suspected patients was 7.2 days. The interval from sample collection to laboratory
receipt was 1.5 days, 1.8 days, and 2.1 days in Guinea, Liberia, and Sierra Leone, respectively
]. These relatively long intervals from symptom onset to diagnosis indicate that early
detection of EVD is hindered by technical and social factors [
]. These delays likely contributed to
the early expansion of the epidemic. Reducing them should improve transmission interruption,
contact tracing, epidemiological surveillance accuracy, and prognoses.
Taking timely measures to control and prevent infection during small outbreaks is
important for maintaining local control and preventing larger outbreaks. Previous studies have
shown that strict isolation and quarantine, scientific management of corpses, public health
education knowledge, and good personal hygiene habits are effective means of controlling and
preventing the perpetuation of infectious diseases [
]. Similarly, we found that the
outbreak and spread of EVD could be controlled effectively through isolating patients early,
quickly removing corpses, practicing “safe” sex after recovery, and strictly monitoring for
imported cases. EVD patients were first reported in Guinea and then rapidly spread to
neighboring countries that failed to prevent and manage imported cases effectively . Preventing
imported cases is a major way that countries can protect themselves from infectious disease.
However, if imported cases do reach a country, an epidemic can still be prevented by rapidly
implementing control and prevention measures. For example, several countries, including the
USA and Spain, had imported EVD cases [
], but outbreaks were controlled successfully due
to the timely isolation of cases and careful monitoring of close patient contacts.
One of the limitations of this study is that the epidemic data used were published by the
WHO and were likely underreported due to the outdated public health system in West African
countries. However, the WHO data are the most accurate data available at present. The
9 / 17
constructed model fit and forecast epidemic trends accurately, evaluated the effects of control
and prevention measures, and provided a scientific conclusion. In the SEIR model constructed
during the 1995 EVD outbreak, the values of the model’s parameters were estimated based on
stochastic process theory. However, in this study, the values of the model’s parameters were
determined based on references or were simulated using real data from the 2014–2015 EVD
outbreak. Therefore, the results of this study would be expected to more reliable because the
model conformed more closely to the real-world circumstances.
In summary, this study indicated that EVD would be expected to re-emerge within two
decades if control and prevention measures are not taken. To avoid similar EVD-related public
health emergencies in the future, it is crucial that EVD vaccines are developed and approved
for distribution in mass vaccination campaigns in Ebola virus-endemic areas. When future
EVD cases are detected, it will be important for healthcare personnel to (1) isolate affected
patients early, (2) trace and isolate infected patients’ close contacts quickly, (3) remove corpses
with minimal delay, (4) conduct strict monitoring for imported cases, and (5) impose on
recovered patients the importance of having condom-protected sex.
Estimation of the effective vaccination rate
The basic reproduction number (R0) was used to estimate the effective vaccination rate in the
whole population. R0 represents the average number of secondary cases caused by an infected
individual throughout the course of infection in a completely susceptible population and in the
absence of control interventions. When R0<1, the disease epidemic will decline; when R0>1,
disease spread will increase [
]. When random and continuous vaccination strategies for
susceptible people were undertaken, the number of infected persons in the next generation will be
less than 1 if at least R0−1 infected persons of the next generation were vaccinated, resulting in
a termination of a pandemic. Therefore, the minimum vaccination rate of the population can
be expressed as follow:
A detailed derivation of the above formula is published elsewhere [
Epidemic trend without control measures
The epidemiology of EVD in the absence of control and prevention measures was analyzed by
a SIR model. The differential equations were as follows:
S0 ¼ bN
I0 ¼ bIS=N
R0 ¼ gI
N0 ¼ ðb
ða þ d þ gÞI
Because the birth rate (b) is greater than mortality rate (d) in West Africa, the population in
this region is expected to increase and the number of individuals in the S, I, and R blocks
should increase as well. Therefore, there was no dynamic equilibrium point in the system. The
following normalization transformation was applied:
s ¼ S=N; i ¼ I=N; r ¼ R=N
The first three equations of equation group Eq (1) can be transformed as follows:
The expression of the modified reproductive number (designated by θ) was:
s þ i þ r
If θ 1, then it can be proved that s, i, and r have a single dynamic equilibrium point of
P0(1,0,0), and the system is asymptotically stable. If θ>1, then P0 is not stable, there is a single
positive dynamic equilibrium point of P (s ,i ,r ), and the system is asymptotically stable [
The calculation result (θ = 1.85), when the three countries are pooled into a single region,
indicates the presence of a positive dynamic equilibrium point.
Next, the net growth threshold of the population was considered to analyze the trend of the
total population N when b>d and θ>1. The net growth threshold of the population (designated
by φ) can be expressed as:
φ ¼ aðs
1Þ ð1 þ dÞ
s ¼ ða þ d þ gÞ
If φ>1 and t ! 1, then N(t) ! 1; if φ = 1, then N(t) !N (finite number); and if φ<1,
then N(t) ! 0 [
]. We estimated the epidemic trend of s, i, and r at the positive dynamic
equilibrium point by applying the Jacobian matrix of equation group Eq (2) and then calculating
the eigenvalue and feature vector. If the eigenvalue is a complex number, then the epidemic is
in periodical oscillation.
Evaluation of the control and prevention measures in the EVD outbreak
A compartment transfer block diagram was constructed based on the transmission patterns
and control measures of EVD (Fig 6B). When a susceptible person in a free environment (S)
comes in contact with a patient in a free environment (I), an un-decontaminated corpse of an
EVD patient (D), or a recovering but still infectious patient (R), some people in S status become
infected and enter the incubation period (E). People in E status can transfer to I or U status
(suspected cases) to accept treatment and be isolated. Some people in the S block who are not
infected, but present with EVD-like symptoms are transferred to the U block. Once a diagnosis
is confirmed in suspected cases in the U block, these confirmed cases and some patients in the I
block are both transferred to the P block (confirmed cases in isolation). Medical staff (H) are at
high risk of infection due to close contact with people in the U and P blocks. Once an individual
is confirmed to be in the I block, their close contacts are immediately monitored and isolated
for observation (Q block). Some infected cases in the Q block are transferred to the P block
after diagnosis. Uninfected cases return to the S status after the maximum incubation period.
11 / 17
Fig 6. Compartment transfer block diagram of the transmission dynamic model. (A) Compartment
transfer block diagram of the SIR model. Blocks S, I, and R represent susceptible individuals, symptomatic
patients, and people with immunity owing to recovery, respectively. β is the standard contact rate, indicating
the number of people infected by the same patient per unit time in a population of entirely susceptible
persons. α and γ denote the probability of one patient dying or recovering per unit time, respectively. b and d
are the population’s birth and mortality rates per unit time, respectively. (B) Comprehensive compartment
transfer block diagram of EVD epidemiology. Blocks S, E, I, and D represent susceptible persons in a free
environment, infected individuals in the incubation period, patients in a free environment, and
not-yetdecontaminated corpses of EVD patients, respectively. Blocks U, P, and Q denote suspected cases in
isolation, confirmed cases in isolation, and close contacts in isolation, respectively. Blocks H, R, K, and A
indicate the medical staff in charge of U and P, recovered patients that are still infectious, recovered
individuals that are not infectious and with immunity, and imported cases per day, respectively.
After treatment, some patients in the P block die and their corpses are rendered harmless by
decontamination; however, other patients in the P block are transferred to the R block
(recovered patients that are still infectious).
Differential equations were built according to the block diagram, as follows:
S0 ¼ A1 þ dqsQ þ dusU f ðI; D; RÞ φðI; UÞ
E0 ¼ A2 þ f ðI; D; RÞ
I0 ¼ A3 þ o1E
12 / 17
D0 ¼ m1I
U 0 ¼ deuE þ φðI; UÞ
P0 ¼ dipI þ dupU þ o2Q þ gðP; U Þ
H0 ¼ A4
gðP; U Þ
Q0 ¼ ðIÞ
R0 ¼ a2P
K0 ¼ gR
f ðI; D; RÞ ¼ bIð1
dipÞI þ bDD þ bRR
φðI; U Þ ¼ kðdipI þ dupU Þ
ðIÞ ¼ 1=kbIdipI
gðP; U Þ ¼ bPðP þ lU Þ
After a confirmed case is identified in a free environment, his/her close contacts are
checked within 24 hours and placed under observation in isolation. Therefore, the term δipβII
is subtracted from f(I,D,R). κ denotes the number of new suspected cases (non-EVD
infection) when a new confirmed case is reported (excluding confirmed cases derived from the
medical staff and close contacts). λ denotes the proportion of confirmed cases in the U block
population. k represents the probability of a susceptible person being infected through close
contact with a patient with EVD. All of these analyses were conducted in Matlab R2012a
(MathWorks, USA, 2012).
The values of the parameters in the models shown in Fig 6A and 6B were obtained in three
ways. The first way involved consultation of references, and the official websites of the WHO
and the United Nations (noted as Reference in Table 1). For parameters that could not be
obtained directly from published data, the second way, namely establishing the best fit model
was used. The nonlinear least-squares method was used to achieve optimized curve-fitting in
Matlab software. The values of parameters could be determined when the residual sum of
squares of the simulation value and the WHO reported value was minimized (noted as
nonlinear least- squares in Table 1). The third way was simple arithmetic calculations based on
the values of parameters obtained from the former two approaches (noted as calculation in
Sensitivity analysis of the comprehensive compartment
PRCC determination combined with Latin hypercube sampling (LHS) was used for the
sensitivity analysis. The PRCC is an efficient and reliable sampling-based sensitivity analysis
method that provides a measure of monotonicity between a set of parameters and the model
output after removing the linear effects of all parameters except the parameter of interest [
13 / 17
LHS is a stratified Monte Carlo sampling method in which each parameter’s range is divided
into N equal intervals and one sample is selected randomly from each interval [
standard correlation coefficient-ρ for the parameter and model output was calculated.
S1 Table. The No. of cumulative cases and deaths published by WHO.
Conceived and designed the experiments: ZG DL ZW. Performed the experiments: ZG DX.
Analyzed the data: ZG. Wrote the paper: ZG DX. Conducted the literature review and collected
datas: XW YW TY ZW. Organized and coordinated all participants to accomplish the study:
15 / 17
16 / 17
1. Cenciarelli O , Pietropaoli S , Malizia A , Carestia M , D'Amico F , Sassolini A , et al. Ebola virus disease 2013 -2014 outbreak in west Africa: an analysis of the epidemic spread and response . Int J Microbiol . In press 2015 .
2. Carroll MW , Matthews DA , Hiscox JA , Elmore MJ , Pollakis G , Rambaut A , et al. Temporal and spatial analysis of the 2014-2015 Ebola virus outbreak in West Africa . Nature . In press 2015 .
3. WHO. Ebola Response Team . Ebola virus disease in West Africa-the first 9 months of the epidemic and forward projections . N Engl J Med . 2014 ; 371 ( 16 ): 1481 - 1495 . doi: 10 .1056/NEJMoa1411100 PMID: 25244186
4. Van Kerkhove MD , Bento AI , Mills HL , Ferguson NM , Donnelly CA. A review of epidemiological parameters from Ebola outbreaks to inform early public health decision-making. Sci Data . In press 2015 .
5. Fauci AS . Ebola-underscoring the global disparities in health care resources . N Engl J Med . 2014 ; 371 ( 12 ): 1084 - 1086 . doi: 10 .1056/NEJMp1409494 PMID: 25119491
6. Chertow DS , Kleine C , Edwards JK , Scaini R , Giuliani R , Sprecher A . Ebola virus disease in West Africa-clinical manifestations and management . N Engl J Med . 2014 ; 371 ( 22 ): 2054 - 2057 . doi: 10 . 1056/NEJMp1413084 PMID: 25372854
7. Matua GA , Van DM , Locsin RC . Ebola virus and haemorrhagic syndrome . Sultan Qaboos Univ Med J . 2015 ; 15 ( 2 ): E171 - 176 . PMID: 26052448
8. WHO. Ebola Response Team. West African Ebola epidemic after one year-slowing but not yet under control . N Engl J Med . 2015 ; 372 ( 6 ): 584 - 587 . doi: 10 .1056/NEJMc1414992 PMID: 25539446
9. Baize S , Pannetier D , Oestereich L , Rieger T , Koivogui L , Magassouba N , et al. Emergence of Zaire Ebola Virus Disease in Guinea-preliminary report . N Engl J Med . 2014 ; 371 ( 15 ): 1418 - 1425 . doi: 10 . 1056/NEJMoa1404505 PMID: 24738640
10. Breman JG , Johnson KM . Ebola then and now . N Engl J Med . 2014 ; 371 ( 18 ): 1663 - 1666 . doi: 10 . 1056/NEJMp1410540 PMID: 25207624
11. Farrar JJ , Piot P. The Ebola emergency-immediate action, ongoing strategy . N Engl J Med . 2014 ; 371 ( 16 ): 1545 - 1546 . doi: 10 .1056/NEJMe1411471 PMID: 25244185
12. Kouadio KI , Clement P , Bolongei J , Tamba A , Gasasira AN , Warsame A , et al. Epidemiological and surveillance response to Ebola Virus Disease outbreak in Lofa County , Liberia (March-September , 2014 ); Lessons Learned . PLoS Curr . In press 2015 .
13. Siettos C , Anastassopoulou C , Russo L , Grigoras C , Mylonakis E. Modeling the 2014 Ebola virus epidemic-agent-based simulations, temporal analysis and future predictions for Liberia and Sierra Leone . PLoS Curr . In press 2015 .
14. McLeod RG , Brewster JF , Gumel AB , Slonowsky DA . Sensitivity and uncertainty analyses for a SARS model with time-varying inputs and outputs . Math Biosci Eng . 2006 ; 3 ( 3 ): 527 - 544 . PMID: 20210378
15. Chowell G , Castillo-Chavez C , Fenimore PW , Kribs-Zaleta CM , Arriola L , Hyman JM . Model parameters and outbreak control for SARS . Emerg Infect Dis . 2004 ; 10 ( 7 ): 1258 - 1263 . PMID: 15324546
16. Sitong L , Litao H , Hongyan L , Zhi D , Qian T , Kaveh K , et al. Evaluating the impact of test-and-treat on the HIV epidemic among MSM in China using a mathematical model . PLoS One . 2015 ; 10 ( 6 ): E0126893. doi: 10 .1371/journal.pone. 0126893 PMID: 26039075
17. Seidu B , Makinde OD , Seini IY . Mathematical analysis of the effects of HIV-Malaria co-infection on workplace productivity . Acta Biotheor . 2015 ; 63 ( 2 ): 151 - 182 . doi: 10 .1007/s10441-015-9255-y PMID: 25980477
18. Endo A , Nishiura H . Transmission dynamics of vivax malaria in the republic of Korea: Effectiveness of anti-malarial mass chemoprophylaxis . J Theor Biol . 2015 ; 380 ( Suppl 1 ): 499 - 505 .
19. Lekone PE , Finkenstädt BF . Statistical inference in a stochastic epidemic SEIR model with control intervention: Ebola as a case study . Biometrics . 2006 ; 62 ( 4 ): 1170 - 1177 . PMID: 17156292
20. World Health Organization. Ebola Situation Reports. Available: http://apps.who.int/Ebola/Ebolasituation-reports. Published Apr 1 , 2015 . Accessed Jul 15 2015 .
21. Christie A , Davies-Wayne GJ , Cordier-Lasalle T , Blackley DJ , Laney AS , Williams DE , et al. Possible sexual transmission of Ebola virus-liberia . Morbidity and Mortality Weekly Report . 2015 ; 64 ( 17 ): 479 - 481 . PMID: 25950255
22. Althaus CL . Estimating the reproduction number of Ebola virus (EBOV) during the 2014 outbreak in West Africa . PLOS Curr . In press 2015 .
23. Fisman D , Tuite A . Projected impact of vaccination timing and dose availability on the course of the 2014 West African Ebola epidemic . PLoS Curr . In press 2015 .
24. Luo T , Sumi A , Zhou D , Kamo K , Yu B , Zhao D , et al. Study on the effect of measles control programmes on periodic structures of disease epidemics in a large Chinese city . Epidemiol Infect . 2011 ; 139 ( 2 ): 257 - 264 . doi: 10 .1017/S0950268810001056 PMID: 20478086
25. Deguen S , Flahault A. Impact on immunization of seasonal cycle of chickenpox . Eur J Epidemiol . 2000 ; 16 ( 12 ): 1177 - 1181 . PMID: 11484809
26. Chua AC , Cunningham J , Moussy F , Perkins MD , Formenty P. The case for improved diagnostic tools to control Ebola virus disease in West Africa and how to get there . PLoS Negl Trop Dis . 2015 ; 9(6): E0003734 . doi: 10 .1371/journal.pntd. 0003734 PMID: 26068890
27. Abhishek P , Katherine EA , Jan M , Natasha W , Jeffrey PT , James EC , et al. Strategies for containing Ebola in West Africa . Science . 2014 ; 346 ( 6212 ): 991 - 995 . doi: 10 .1126/science.1260612 PMID: 25414312
28. Fast SM , Mekaru S , Brownstein JS , Postlethwaite TA , Markuzon N. The role of social mobilization in controlling Ebola virus in Lofa County, Liberia . PLoS Curr . In press 2015 .
29. WHO. Ebola response roadmap . 2015 ; 7 : 15 . Available: http://apps.who.int/iris/bitstream/10665/ 136508/1/roadmapsitrep. Published Oct 15, 2014 .
30. Diekmann O , Heesterbeek J . Mathematical Epidemiology of Infectious Diseases: Model Building. 1st ed. West Sussex , England: John Wiley & Sons Ltd Press; 2000 .
31. Chauhan A , Misra OP , Dhar J . Stability Analysis of SIR Model with Vaccination . American Journal of Computational and Applied Mathematics . 2014 ; 4 ( 1 ): 17 - 23 .
32. Busenderg S , Driessche PV . Analysis of a disease transmission model in a population with varying size . J Math Biol . 1990 ; 28 ( 3 ): 256 - 270 .
33. United Nations , Department of Economic and Social Affairs. Rate of natural increase by major area, region and country, 1950-2100 (per 1,000 population) . Available: http://esa.un.org/unpd/wpp/DVD/#. Published July 2015 .
34. United Nations , Department of Economic and Social Affairs. Crude death rate by major area, region and country, 1950-2100 (deaths per 1,000 population) . Available: http://esa.un.org/unpd/wpp/DVD/#. Published July 2015 .
35. United Nations , Department of Economic and Social Affairs. Total population (both sexes combined) by major area, and country, annually for 1950-2100 (thousands ). Available: http://esa.un.org/unpd/wpp/ DVD/#. Published July 2015 .
36. Marino S , Hogue IB , Ray CJ , Kirschner DE . A methodology for performing global uncertainty and sensitivity analysis in systems biology . J Theor Biol . 2008 ; 254 ( 1 ): 178 - 196 . doi: 10 .1016/j.jtbi. 2008 . 04 .011 PMID: 18572196
37. McKay MD , Beckman RJ , Conover WJ . A comparison of three methods for selecting values of input variables in the analysis of output from acomputer code . Technometrics . 1979 ; 21 ( 2 ): 239 - 245 .