A model for a chikungunya outbreak in a rural Cambodian setting: implications for disease control in uninfected areas

Abstract

Following almost 30 years of relative silence, chikungunya fever reemerged in Kenya in 2004. It subsequently spread to the islands of the Indian Ocean, reaching Southeast Asia in 2006. The virus was first detected in Cambodia in 2011 and a large outbreak occurred in the village of Trapeang Roka Kampong Speu Province in March 2012, in which 44% of the villagers had a recent infection biologically confirmed. The epidemic curve was constructed from the number of biologically-confirmed CHIKV cases per day determined from the date of fever onset, which was self-reported during a data collection campaign conducted in the village after the outbreak. All individuals participating in the campaign had infections confirmed by laboratory analysis, allowing for the identification of asymptomatic cases and those with an unreported date of fever onset. We develop a stochastic model explicitly including such cases, all of whom do not appear on the epidemic curve. We estimate the basic reproduction number of the outbreak to be 6.46 (95% C.I. [6.24, 6.78]). We show that this estimate is particularly sensitive to changes in the biting rate and mosquito longevity. Our model also indicates that the infection was more widespread within the population on the reported epidemic start date. We show that the exclusion of asymptomatic cases and cases with undocumented onset dates can lead to an underestimation of the reproduction number which, in turn, could negatively impact control strategies implemented by public health authorities. We highlight the need for properly documenting newly emerging pathogens in immunologically naive populations and the importance of identifying the route of disease introduction.

Figure 1. Epidemic curve showing confirmed chikungunya per day by date of reported onset in the village of Trapeang Roka, Cambodia.

The grey arrow indicates the start of a two-day rain spell.

Figure 2. A map of Trapeang Roka village, showing all houses for which gps co-ordinates were collected.

The map shows the distribution of biologically-confirmed symptomatic cases, documented by date of fever onset. Each circle denotes the location of a house within the village. Unfilled circles indicated houses that escaped infection. The colour bar indicates the number of symptomatic cases with a documented date of symptom onset in each house. The black diagonal line indicates the main road running through the village, about which the houses are clustered.

Figure 3. A map of Trapeang Roka village, showing all houses for which gps co-ordinates were collected.

The map shows houses with no confirmed infection (unfilled circle), houses with only infections documented by date of onset (black circle), houses with only infections undocumented by date of onset (red circles) and houses which have both cases with documented and undocumented infection onset dates (green circle). The black diagonal line indicates the main road running through the village, about which the houses are clustered.

Figure 4. Bar charts showing the distribution of cases with documented and undocumented (including asymptomatic) dates of symptom onset within the population.

(a) gender of cases, (b) age group of cases in years, (c) education level of cases: No schooling (N), Primary school (P), Secondary school (S), (d) occupation of cases: Student (S), Stay at home (H), Factory worker (W), Construction worker (C), Child (Ch), Vendor (V), Farmer (F).

Figure 5. Schematic of the disease transmission pathway.

Black arrows indicate transitions between disease states. A susceptible mosquito () can be infected by a symptomatic human documented by date of onset (), a symptomatic human undocumented by date of onset () or an asymptomatic () human (dashed blue arrow). A susceptible human () can be infected by an infected mosquito () (dashed red arrow).

Figure 6. The mean of 1000 stochastic realisations for the number of daily symptomatic cases documented by date of onset (solid black line) plotted with the epidemic curve (solid red line).

Also shown is the mean of 1000 realisations of the SEIR model (dashed black line) and an eigendecomposition of the epidemic curve (dashed red line). The grey shaded area shows the 95% confidence interval. Day 0 corresponds to the start of the epidemic on February 7.

Figure 7. The mean of 1000 stochastic realisations for number of symptomatic cases documented by date of onset (solid black line), symptomatic cases undocumented by date of onset (dotted black line), asymptomatic cases (dashed black line) and the total number of infectious cases (solid blue line) plotted with the epidemic curve (solid red line).

Day 0 corresponds to the start of the epidemic on February 7.

Figure 8. Tornado diagram of univariate sensitivity analysis.

The diagram shows the degree to which a 10% variability in the parameters affects the value of . Each bar is a representation of how uncertainty in that particular parameter affects the estimate of the reproduction number. The baseline scenario is fixed with .