Utility of surveillance data for planning for dengue elimination in Yogyakarta, Indonesia: a scenario-tree modelling approach

Abstract

Introduction: Field trials and modelling studies suggest that elimination of dengue transmission may be possible through widespread release of Aedes aegypti mosquitoes infected with the insect bacterium Wolbachia pipientis (wMel strain), in conjunction with routine dengue control activities. This study aimed to develop a modelling framework to guide planning for the potential elimination of locally acquired dengue in Yogyakarta, a city of almost 400 000 people in Java, Indonesia.

Methods: A scenario-tree modelling approach was used to estimate the sensitivity of the dengue surveillance system (including routine hospital-based reporting and primary-care-based enhanced surveillance), and time required to demonstrate elimination of locally acquired dengue in Yogyakarta city, assuming the detected incidence of dengue decreases to zero in the future. Age and gender were included as risk factors for dengue, and detection nodes included the probability of seeking care, probability of sample collection and testing, diagnostic test sensitivity and probability of case notification. Parameter distributions were derived from health system data or estimated by expert opinion. Alternative simulations were defined based on changes to key parameter values, separately and in combination.

Results: For the default simulation, median surveillance system sensitivity was 0.131 (95% PI 0.111 to 0.152) per month. Median confidence in dengue elimination reached 80% after a minimum of 13 months of zero detected dengue cases and 90% confidence after 25 months, across different scenarios. The alternative simulations investigated produced relatively small changes in median system sensitivity and time to elimination.

Conclusion: This study suggests that with a combination of hospital-based surveillance and enhanced clinic-based surveillance for dengue, an acceptable level of confidence (80% probability) in the elimination of locally acquired dengue can be reached within 2 years. Increasing the surveillance system sensitivity could shorten the time to first ascertainment of elimination of dengue and increase the level of confidence in elimination.

Keywords: Dengue; Mathematical modelling.

Figure 1

Scenario-tree model for the detection of locally acquired, symptomatic, virologically confirmed dengue cases through the public health surveillance system in Yogyakarta city, Indonesia. The scenario-tree model represents the planned structure of the surveillance system in the context of a prospective dengue elimination study. This includes an enhanced clinic-based surveillance component at primary health facilities (puskesmas), as well as strengthening the hospital-based surveillance system through introduction of virological testing for all suspected dengue cases. The rectangle nodes represent risk nodes: the probability of being a symptomatic dengue case varies by age and gender. The oval nodes represent detection nodes: key events, actions, or tests that contribute to an infected individual’s detection in the surveillance process. The triangular nodes represent pathways by which cases escape detection.

Figure 2

Box and whisker plot of component sensitivities for puskesmas-based enhanced dengue surveillance (CSe_puskesmas) and hospital surveillance for dengue (CSe_hospital), and overall surveillance system sensitivity for dengue cases (SSe), per month.

Figure 3

Median confidence of dengue elimination over 60 months, for probability of recurrence (PRecur) of 1%, 2% and 5% per month. PRecur: probability of recurrence of local dengue transmission in a single time period (1 month). Reference lines for level of confidence of elimination are plotted at 80% confidence (green dashed line), and 95% confidence (red reference line).