Evaluating the effectiveness of localized control strategies to curtail chikungunya

Chikungunya, a re-emerging arbovirus transmitted to humans by Aedes aegypti and Ae. albopictus mosquitoes, causes debilitating disease characterized by an acute febrile phase and chronic joint pain. Chikungunya has recently spread to the island of St. Martin and subsequently throughout the Americas. The disease is now affecting 42 countries and territories throughout the Americas. While chikungunya is mainly a tropical disease, the recent introduction and subsequent spread of Ae. albopictus into temperate regions has increased the threat of chikungunya outbreaks beyond the tropics. Given that there are currently no vaccines or treatments for chikungunya, vector control remains the primary measure to curtail transmission. To investigate the effectiveness of a containment strategy that combines disease surveillance, localized vector control and transmission reduction measures, we developed a model of chikungunya transmission dynamics within a large residential neighborhood, explicitly accounting for human and mosquito movement. Our findings indicate that prompt targeted vector control efforts combined with measures to reduce transmission from symptomatic cases to mosquitoes may be highly effective approaches for controlling outbreaks of chikungunya, provided that sufficient detection of chikungunya cases can be achieved.

Scientific RepoRts | 6:23997 | DOI: 10.1038/srep23997 To assess the effectiveness of interventions against chikungunya, we developed a stochastic vector-borne transmission model for the spread of chikungunya within a residential neighborhood. Our model takes into account the impact of household transmission of chikungunya as well as human and mosquito movement between houses. We parameterized the model using epidemiological, entomological, clinical, and human movement data. For both Ae. aegypti and Ae. albopictus, we evaluated the effectiveness of intervention strategies for mitigating the spread of a chikungunya outbreak. In contrast to the contention that little can be done to contain chikungunya 11 , we found that vector control and reducing contact between symptomatic humans and mosquitoes can interact synergistically to effectively contain chikungunya. Specifically, we found perifocal mosquito control targeting residences of infected individuals to be essential in reducing mosquito infection, while effective transmission reduction measures of patients after diagnosis should prevent further infection. Implemented together, these interventions can substantially reduce the attack rate of chikungunya and contain outbreaks.

Methods
To simulate viral transmission in a residential neighborhood, we developed a household-level spatial stochastic model of chikungunya transmission, explicitly modeling individual human and mosquito movement and exposure. We ran our continuous-time model for one year using the Gibson-Bruck adaptation of the Gillespie algorithm 26 . Using this algorithm, we iteratively sampled and executed randomly chosen events, including entomological events, such as the movement of a mosquito to an adjacent house, or epidemiological events, such as the acquisition or clearance of chikungunya infection by a human. After each event was executed in the simulation, the transition probabilities for all events were updated to reflect the new state of the system. We ran our model on a two-dimensional square lattice with 20 × 20 sites, representing a neighborhood of 400 houses populated with four residents per house 27,28 . As our time scale of interest was a one-year period, we assumed the human population to be constant with no human mortality, and the mosquito population dynamics to be governed by a household-level carrying capacity.
We assumed that humans divide their time between their homes and other houses, with the likelihood of the latter depending on the distance of other residences away from their homes, as informed by an empirical study 29 . Each house i was assigned a number of contact houses n i sampled from an exponential distribution with mean of 6.26 29 . We assumed that, on average, an individual in house i spends a fraction τ = .
. 30 of time at home. The remaining time was spent in one of the n i contact houses, chosen randomly and each assigned an exponentially distributed weight τ ij and normalized such that Individuals were designated as susceptible, exposed, infected (pre-symptomatic), infected (70-90% of whom were asymptomatic), or recovered from chikungunya, at rates parameterized from epidemiological, clinical and entomological studies ( Table 1). The transmission of chikungunya from infectious mosquitoes in house j to humans from house i was modeled as λ human, c is the biting rate, I v j is the number of infectious mosquitoes in house j, S h i is the number of susceptible humans in house i, N hi is the number of humans in house i, and τ ij is the proportion of time spent in house j by the individual from house i. After an incubation period ranging from two to six days, exposed humans become pre-symptomatically infectious for two days 3,24,31,32 . Following this pre-symptomatic period, a proportion κ of infectious humans become symptomatic while the rest remain asymptomatic.
Mosquito dynamics and epidemiology. We parameterized the daily rate of mosquito movement between houses from a field study (r m = 0.044 per day) 33 . We specified a household-level carrying capacity for each house, and sampled the carrying capacity independently for each house from an exponential distribution of mean 7.8 mosquitoes 34 . For simplicity, we assumed periodic boundary conditions for mosquito movement. In the absence of interventions, humans were bitten by mosquitoes at a fixed daily rate within every home [35][36][37][38] . As described below, we modeled transmission-reduction measures, such as bed nets, in terms of a reduced biting rate. We accounted both for the bites on infectious human hosts that could result in infection of a susceptible mosquito and, conversely, for the bites from infectious mosquitoes that could infect susceptible humans.
Each mosquito was designated as susceptible, exposed, or infected with chikungunya. The transmission of chikungunya from asymptomatic infectious humans residing in house i to mosquitoes in house j was given as where β h is the transmission rate from human to mosquitoes, S v j is the number of susceptible mosquitoes in house j,  I h i is the number of pre-symptomatic infectious humans from house i, and A h i is the number of asymptomatic infectious humans from house i. Based on empirical data of humans symptomatic with chikungunya, we assumed a 50-80% reduction in mobility of symptomatic infected humans 23,39,40 . The corresponding transmission from symptomatic humans in house i to mosquitoes in house j was given as , where χ is the reduction in mobility of symptomatic individuals and I sh i is the number of symptomatic infectious humans from house i. Exposed mosquitoes become infectious after an extrinsic incubation period and remain infectious until they die. We separately considered Ae. aegypti and Ae. albopictus mosquito populations, parameterized with respective epidemiological and entomological data (Table 1). We provided a model diagram for illustration ( Fig. 1).

Calibration.
We evaluated the effectiveness of containment intervention strategies for mitigating the spread of an emerging chikungunya outbreak. We used the basic reproduction number (R 0 ) as a measure of the likelihood of disease outbreak and intensity of spread through the population. To calculate R 0 , we simulated the number of secondary cases of human infections that resulted from mosquitoes that were directly infected by an initial human case in a population of fully susceptible humans and mosquitoes. By simulating 10,000 iterations of our stochastic model, we generated a probabilistic distribution of secondary cases likely to arise from a single infection, with mean R 0 41,42 . The transmission parameter β v was separately calibrated for both Ae. aegypti and Ae. albopictus by fitting the simulated distribution of R 0 against empirical estimates from previous and current chikungunya outbreaks 31,43 . These chikungunya  43 . With the exception of β v , which we calibrated, the model parameters were obtained from literature (Table 1). We illustrate this method of calculation in Supplementary Material (Fig. S1).
Containment measures for chikungunya include perifocal space spraying (fogging) of insecticides in perimeter radius around houses with detected cases, indoor residual spraying 22,45,46 , the application of mosquito repellents on symptomatic individuals and the use of bed nets for patients in their homes 3,45,46 . We considered intervention strategies combining different levels of: 1) perifocal and indoor residual spraying in houses with reported cases and directly neighboring houses and 2) transmission reduction measures, specifically bed nets and/or mosquito repellents, for diagnosed individuals to reduce subsequent mosquito bites, both to household members and neighbors. We assumed that vector control only affected mosquitoes that were present in and around treated houses, with the efficacy of vector control incorporated into the model as the probability of mosquito mortality from perifocal or indoor residual spraying. The efficacy of transmission reduction measures for symptomatic individuals was incorporated by reducing the contact rate between humans and mosquitoes which results in a reduction of transmission rates of human-to-mosquito and mosquito-to-human. With the implementation of these intervention measures, the transmission from symptomatic humans to mosquitoes was given as i , and q is their efficacy for reducing contact with mosquitoes. We evaluated the effectiveness of intervention strategies for mitigating a chikungunya outbreak over varying levels of disease surveillance, defined as the probability of case detection. Effectiveness of interventions was measured in terms of the attack ratio, defined as the proportion of individuals who are infected over the course of the epidemic. The attack ratio was computed by averaging the proportion of individuals infected during the simulated one year epidemic period over the 1000 iterations of our model. To account for differing efficacies of and compliance to the control measures, we varied the efficacy of the intervention strategy for detecting new cases of chikungunya, removing adult mosquitoes through vector control, and reducing biting rate on symptomatic individuals to whom transmission reduction measures are targeted. We considered different levels of chikungunya transmissibility (R 0 = 2, 4, 6), efficacies of vector control for killing adults mosquitoes (ranging from 10 to 90%), and sensitivities of surveillance measures for detecting new cases of symptomatic human infection (low (30% efficacy), intermediate (50% efficacy), high (80% efficacy)). A flowchart outlines the steps of our analysis (Fig. S2).

Results
For Ae. aegypti, the transmission parameter β v was estimated to be 0.06 when R 0 equals 2, 0.14 when R 0 equals 4 and 0.24 when R 0 equals 6. For Ae. albopictus, the transmission parameter β v was estimated to be 0.1 for R 0 equals 2, 0.22 for R 0 equals 4 and 0.36 for R 0 equals 6. We validated the time series dynamics of our model against empirical data from the 2013-2014 chikungunya outbreak in Dominica (Fig. 2), and illustrated the spatiotemporal dynamics with spatial snapshots of cases throughout the modelled residential neighborhood (Fig. 3).
We evaluated the effectiveness of a reactive perifocal vector control strategy for curtailing chikungunya, which consists of insecticide spraying in perimeter radius around houses with infected individuals, as well as indoor residual spraying. We found that perifocal vector control has the potential to be effective for curtailing the spread of a chikungunya outbreak. However, its effectiveness is sensitive to the efficacy of surveillance measures for identifying new cases (Fig. 4). We showed that for a low 30% probability of detecting new symptomatic cases, even a highly efficacious vector control measure with 90% efficacy would have only a marginal reduction of the epidemic attack ratio (Fig. 4). However, for a detection efficacy of 80%, vector control has the potential to substantially reduce the epidemic attack ratio even when the vector control has an intermediate efficacy of 60% (Fig. 4).
We evaluated the effectiveness in terms of reducing the attack ratio of chikungunya of an intervention strategy combining perifocal vector control with measures that decrease mosquito biting of infectious individuals. We found that such transmission reduction measures interact synergistically with vector control to greatly improve the effectiveness of transmission mitigation, especially when vector control has low efficacy (Figs 4 and 5). In low transmission intensity settings (e.g. R 0 = 2), the combination of disease surveillance measures with at least 80% probability of detecting new cases, efficacious transmission reduction measures for symptomatic individuals with at least 80% reduction of contact between infected individuals and mosquitoes, and vector control measures with at least 50% efficacy in killing adult mosquitoes was found to reduce the epidemic attack ratio from 20%, without control, to less than 1% (Figs 4 and 5). In high transmission intensity settings (e.g. R 0 = 6), the combined intervention was found to reduce the epidemic attack ratio from 60% down below 20% (Figs 4 and 5).
We also evaluated the impact of the combined intervention strategies on reducing the average number of secondary cases of human infections resulting from mosquitoes that were directly infected by the initial human case. We found that control measures combining surveillance, transmission reduction for symptomatic individuals, and perifocal vector control at an efficacy of at least 80% reduced initial R 0 values of 3 − 6 to below 2, and initial R 0 values of 3 and lower to below 1 (Fig. 6).
In our uncertainty analysis, we considered the impact of containment measures and of delays in implementing interventions upon chikungunya transmission. We evaluated the probability of preventing an outbreak, defined as the probability that an index case would not lead to less than one secondary case, when vector control and transmission reduction measures are implemented (Fig. 7A,C). We found that, even with a two week delay in initiating intervention, integrated efficacious vector control and transmission reduction for symptomatic individuals may be able to prevent a chikungunya outbreak (Fig. 7A,C). For example, control measures that combine surveillance with at least 70% case detection efficacy, transmission reduction measures and perifocal vector control, at an efficacy of at least 70% for each intervention, have at least a 60% probability of preventing a chikungunya outbreak when R 0 is below 4 (Fig. 7A,C). However, if R 0 is above 4, control measures fail to prevent an outbreak (Fig. 7A,C), although the magnitude of the resulting epidemic would vary substantially with the efficacy of disease surveillance measures (Fig. 7B,D). The probability of containing an outbreak decreases as both R 0 and delay in initiating interventions increases (Fig. 7).

Discussion
With nearly the entire Western Hemisphere presently at risk for chikungunya invasion, evaluating the effectiveness of measures for containment has become a public health priority 14 . Given that there are currently no vaccines nor specific treatments effective in reducing chikungunya transmission, it has been argued that little can be done to control chikungunya 11 . Using a spatial stochastic model for chikungunya transmission, we showed that, contrary to this assessment, perifocal vector control is capable of limiting the spread of chikungunya. We found that when vector control is integrated with transmission reduction around symptomatic individuals, the combined intervention strategy has the potential to interrupt disease transmission even after accounting for asymptomatic infection. We further found that timely implementation of combined interventions early in the outbreak is paramount. The strategies that we evaluated focus on targeted containment rather than widespread vector control, as has been successfully applied to contain the invasion and spread of dengue viruses in non-endemic areas 47 . Although the impact of perifocal vector control depends on the sensitivity of disease surveillance for detecting new cases of chikungunya, a high degree of accurate and swift clinical diagnosis has been achieved for cases in past epidemics 23,24 .  Previous chikungunya models have investigated the effectiveness of removing mosquito breeding sites 6,48 and chemical control tools, such as adulticide and larvicide 48,49 , for controlling ongoing epidemics. However, these studies assumed homogeneous mixing of human and mosquito populations, and so could not accurately evaluate the effectiveness of reactive control strategies such as perifocal space spraying (fogging) of insecticides and indoor residual spraying 22,45 , which are spatially targeted strategies. Moreover, some of these studies have investigated the impact of control strategies through sensitivity analysis of model input parameters, rather than explicitly quantifying the intensity and speed of control efforts needed for curtailing disease spread 49,50 . Our study extends previous analyses by accounting for human and mosquito movements, focal application and impact of insecticide spray, as well as the integration of early vector control with transmission reduction measures for mitigating the spread of chikungunya. In so doing, our study evaluates optimal speed and efficacy of perifocal vector control and transmission reduction measures needed to curtail a chikungunya outbreak in a large residential neighborhood. Adherence to the use of transmission reduction measures such as bed nets, mosquito repellents, and long-sleeved garments may be promoted through health education that emphasizes the importance of these measures for reducing risk of infection for other household members 51 . Our model is applicable to tropical regions in which the continual abundance of mosquito populations gives chikungunya the potential to circulate throughout the year 52,53 . Climatic variables and other environmental factors may affect mosquito population dynamics and the risk of chikungunya outbreaks, especially in temperate regions 52,53 , where chikungunya transmission may vary seasonally 6,53 . Future study may take into account climatic factors, while the population social structure could be extended to account for age and gender heterogeneity, which may result in differential contributions to chikungunya transmission 54 . Previous studies have suggested that the effectiveness of spatially targeted insecticide spraying strategies for dengue control may be strongly affected by socially structured human movement 30 . Differences in the social structure of human movement, such as between rural versus urban settings, may impact the effectiveness of chikungunya perifocal vector control strategies. For example in this study, on the basis of empirical data from the city of Iquitos, Peru, we assumed that the time spent by individuals in other residences depends on the proximity to home. However, this assumption may not be applicable to all settings. Therefore, fruitful areas for future research include investigating the effectiveness of perifocal vector control for curtailing chikungunya transmission in settings with socially structured human movement different than the one considered here. In addition, future studies may evaluate the impact of vertical virus transmission within the vector population on the spread of chikungunya and the effectiveness of control strategies for curtailing outbreaks, as well as consider a wide range of R 0 values as some future chikungunya strains may have a higher R 0 value than those considered here.
The early identification of a chikungunya epidemic is a fundamental first step towards implementing effective interventions to rapidly control the disease as well as minimize mortality and morbidity in human populations. Figure 7. Impact of the efficacy of case detection on the chikungunya attack ratio when combined with efficacious vector control and transmission reduction around symptomatic individuals. The efficacies of vector control and transmission reduction measures were varied from 70 to 90%. We defined the effectiveness of controlling the outbreak as the probability that an index case will produce fewer than two secondary cases. The attack ratio was computed when an index case produced at least two secondary cases. We compared timing of intervention initiation with a one-week versus two-week delay from detection of the index case.
Scientific RepoRts | 6:23997 | DOI: 10.1038/srep23997 Early epidemic identification depends on the efficacy of early warning systems that requires continual and timely surveillance of potential human and mosquito cases, as well as climate drivers for chikungunya transmission. While the control of chikungunya is challenged by case detection, efficacy of vector control and the risk of disease reintroduction from other locations, our model shows that early case detection, combined with perifocal vector control and transmission reduction measures, may be an effective approach for mitigating outbreaks. These findings suggest that such a proactive, well-targeted approach to the containment of chikungunya would complement the traditional vector control measures commonly employed against dengue and other mosquito-borne diseases [55][56][57] . If chikungunya remains inadequately addressed, the threat that it poses to many regions of the world will likely increase as the mosquito vectors continue to expand geographically.