Considerations for the 2030 Sustainable Development Goals for dengue

Focus on interventions that controls onwards spread

The combination of complex dengue epidemiology and the lack of effective interventions meant that the feasibility of the 2012 WHO goals were always in doubt. The development of new technologies focused on reducing transmission will be central to tackling future morbidity from dengue. The results from modelling suggest that intervention with efficacy ~70% will be required to achieve negligible annual cases (i.e., reduce the reproductive number to under one). Current available interventions have an estimated efficacy far short of this value. We currently do not know whether an intervention (or combination of interventions) with those characteristics will become available in the next decade. Given that current interventions are insufficient to eliminate transmission, even settings that are able to reduce transmission will continue to experience regular instances of ongoing transmission and outbreaks. A goal of developing and rolling out of interventions of a sufficient efficacy to prevent transmission can help motivate efforts on technologies that have the ability to control onwards spread.

Case fatality rate (CFR) definitions need to be clear

A goal focused on reducing mortality should be central to future efforts to reduce the overall burden from dengue. Improvements in case management can lead to substantial reductions in the number of deaths from dengue⁴³. Some countries have been able to reduce mortality to very low levels. For example, Singapore has some of the best clinical care globally and has achieved a CFR of under 0.05%⁴⁴. Understanding reductions in mortality relies on a clear definition of the CFR. Reported CFRs are usually calculated as the proportion of detected hospitalised severe cases that are fatal, however, there is no consistent definition. Radically different estimates of CFR are possible depending on whether the denominator used is number of hospitalised/severe dengue cases or incidence of milder dengue fever (Figure 1). In addition, CFR may vary significantly if confirmed cases or all suspected cases are used. Dengue infection results in a wide spectrum of disease, with severe symptoms only occurring in a minority of infections⁴⁵. The cases that are detected by a surveillance system will depend on healthcare seeking patterns, local clinical management protocols, healthcare resources and surveillance reporting processes². These factors mean that the number of reported cases can vary considerably across settings and over time. Even in well-resourced settings, most dengue cases are based on imperfect clinical diagnoses without confirmatory testing, especially with milder cases.

Figure 1. Case fatality rate (CFR) estimates with different reporting scenarios.

Assumes 40% of infections are symptomatic, 5% are hospitalised, 1% are very severe and 0.001% are fatal. For misclassification bias, assumes 5% of fatal cases and 20% of hospitalised cases are misclassified.

Avoid goals based on reducing dengue ‘outbreaks’

For some pathogens, reducing the frequency and size of epidemics may form the basis for a simple and easy to understand goal to reduce disease burden. For example, the size and duration of outbreaks has been useful in measuring the pathway to measles and rubella elimination⁴⁶. However, the complexities of the dengue disease system severely reduce the potential for this approach here. Firstly, progress to outbreak-related goals cannot be causally associated with interventions. The underlying epidemiology of dengue is complex. In the absence of interventions, there can be orders of magnitude difference in the observed number of cases across years. For example, in Bangkok, Thailand, there are large fluctuations in the number of cases reported each year and continual year-round circulation despite largely consistent behaviour in intervention deployment (Figure 2). This poses challenges to monitoring the progress of reducing transmission risk from interventions.

Figure 2. Observed case counts in a tertiary care hospital in Bangkok, Thailand between 1973 and 2015.

While the number of observed cases has risen steadily over this period, the underlying force of infection has fallen by over 50%¹¹.

Secondly, historic incidence in many locations is poorly characterized, limiting ability to define ‘putative epidemics’ - goals based on outbreaks necessitate a robust definition of what constitutes an ‘outbreak’. Most approaches rely on using background levels of incidence in a location to create epidemic thresholds⁴⁷. The type of cases captured by surveillance systems will differ substantially across locations (by level of severity, confirmation status) meaning there will be inconsistent definitions of epidemics in different countries. Changes in surveillance over time will also complicate the use of historical observed incidence. For example, using models fit to age data (which is more robust than changes in surveillance²⁰), it has been shown that over the period 1980 to 2000 there was a 50% decrease in the force of infection in Bangkok, Thailand¹¹. However, changes in surveillance, healthcare seeking and population growth have contributed to a large increase in detected cases over this period (Figure 2).

Finally, surveillance efficiency scales probability of detecting outbreaks. There is a complex interaction between surveillance and the detection of putative outbreaks. Improved surveillance can accompany improvements in control but can also lead to detection of larger number of putative outbreaks simply because of improved detection. This may create situations where settings with the most effective control may detect larger numbers of outbreaks.

Measurement and tracking progress

Specific definitions will be key to reliably achieving the goals. For the first goal, the numerator and the denominator need to be well defined. Most cases are diagnosed only clinically without systematic confirmatory laboratory testing. The true cause of death (numerator) or whether the patient was actually a dengue case is often unclear. Consistent denominator(s) to define what constitutes a ‘case’ need to be agreed. As the CFR varies by age, countries need to capture the age distribution of cases and deaths. Precise definition is also needed to reduce the risk of perverse incentives. The absence of specific guidance on case definitions could generate perverse incentives to lower measured CFR. For example, including non-specific febrile syndromes in the denominator or, in the absence of systematic confirmatory testing, classifying deaths as not dengue attributable to lower the numerator.

Spatially and temporally resolved estimates of dengue transmission (number of infections) and burden (number of symptomatic cases, severe cases, hospitalised cases, deaths) are lacking. Identifying subnational locations that experience the highest transmission and burden is critical to underpinning the targeted training of healthcare staff, allocation of and any future targeted deployment of new tools that can reduce transmission. Age-specific case data can be used to estimate infection intensity and can be paired with nationally representative seroprevalence studies in random subsets of the population to produce robust estimates of infection risk.

In order to help track the progress towards these goals, countries should implement WHO recommended guidance (to be generated) on measuring cases and deaths to document their progress in reducing CFR. Countries should also report subnational age distribution of cases (number of cases by single age class at admin level 1 or higher) as this allows infection risk estimation that is robust to surveillance bias. Countries should also implement WHO recommended guidance (to be generated) on measuring infection and morbidity (both overall and age-specific) to document progress in targeting interventions to 20% of their populations at risk. The guidance would include recommendations on using clinical cases and serological studies among other methods to measure risk and burden⁴⁸.

Technical feasibility

Reducing CFR to <0.05% with modern standards of care is technically feasible, however, it will depend on the clinical care available. While Singapore has a comprehensive surveillance system and some of the best clinical care and has achieved a very low CFR⁴⁴, other countries have poorer surveillance and absence of notified dengue-related deaths cannot be interpreted as absence of undetected mortality. The feasibility of rolling out an intervention with demonstrated potential to reduce dengue transmission or symptomatic infections by over 70% in all endemic countries will depend on such intervention (or combination of interventions) becoming available.

Operational feasibility

For the first goal, the targeted case management training of healthcare workers will be essential. In countries with limited existing surveillance, it will be particularly necessary to identify areas at risk where healthcare staff should be targeted for training and where resources should be allocated to guarantee adequate diagnosis and management of cases. It will also be necessary to monitor changing patterns of age groups at risk (e.g., whether resources should be concentrated in paediatric vs adult clinics) and the populations at risk (e.g., expanding geographic zones) to identify which locations and which health care workers (e.g., adult vs. paediatric physicians) are in need of training.

To meet the second goal, intervention(s) that reduce disease by 70% will need to be developed and identified. It will also be necessary to monitor uptake of an intervention and its effectiveness by maintaining surveillance for disease. For interventions that reduce burden by reducing transmission, it will be necessary to assess whether intervention effectiveness declines over time because of the reduction of immunity due to natural exposure that will occur with the reduction of infection.

It is likely that the optimal deployment of intervention tools will differ across settings, depending on the level of endemicity, population structure, existing surveillance capabilities and available interventions. This will require customized deployments for each setting.

Considerations of cost

Reducing the case fatality and dengue transmission will likely result in savings over the long term due to reduced healthcare expenditure. However, the precise impact will depend on the characteristics of the tools and their efficacy. The rollout of large-scale interventions in all dengue endemic countries will require major funding commitments