Dengue fever is an airborne disease that follows climate and seasonal patterns. Prediction models can be useful to forecast dengue epidemics by creating an early warning system to help guide prevention programmes, such as mosquito control
and public health campaigns.
Dengue is a viral infection transmitted through the bite of an Aedesaegypti mosquito, most commonly found in tropical and subtropical countries. Dengue is a climate-driven disease and highly sensitive to changes in
temperature, rainfall, humidity and wind.
Temperature and rainfall have been shown to be important factors in disease infectiousness and transmission, by affecting the development and lifespan of mosquitoes. In the last decade, average global temperatures have increased by
approximately 0.15°C to 0.2°C per decade, impacting dengue epidemiology in endemic countries.
Humid environments are ideal for mosquito egg survival (over 95% humidity). Moreover, the increase of urbanisation plays an important role in epidemics. Mosquitoes prefer to lay their eggs in water containers inside homes, car tyres and
anywhere where rainwater might collect.
However, any variation in the climate can delay dengue incidence by one to two months. Previous studies that have focused on the relationship between climate change and dengue have shown that long-term climate data and comprehensive
quantitative modelling are important for epidemic preparedness.
Vietnam’s plan to forecast dengue epidemics
The disease burden of dengue fever is significant. In Vietnam, around two million people get infected yearly. However, these rates are suspected to be underreported, as many may not seek healthcare assistance and in some cases may even be
asymptomatic. These factors, combined with changing climate conditions, cause dengue rates to be highly variable.
Therefore, the need for a sophisticated warning system for Vietnam and many other tropical countries would be useful for public health and help build sustainable prevention methods.
‘The disease burden of dengue fever is significant. In Vietnam, around two million people get infected yearly.’
Previous studies have highlighted the usefulness of climate-driven epidemiological surveillance, such as sub-seasonal forecasting, which can predict epidemics between two weeks and two months ahead. In a similar case, Uganda deployed a
forecasting programme created in association with the European Centre for Medium-Range Weather Forecasts that used temperature and rainfall to predict Malaria transmissions four months ahead.
Some of the obstacles in predicting epidemics have been the lack of detailed healthcare records and climate data outputs. Researchers from Vietnam believe that using a sophisticated superensemble model, an algorithm that uses statistical
regression to measure an association between multiple factors, might be the answer. The spatiotemporal (place and time) probabilistic model combines past healthcare records of outbreaks with Earth observation data to predict future
outbreaks at specific locations and times of the year.
The model was designed in partnership with the World Health Organization (WHO), the United Nations Development Programme (UNDP), the Vietnamese Ministry of Health and other Vietnamese health entities. The objective is to estimate the
monthly cases of dengue across the six provinces of Vietnam in real time.
The project collated dengue surveillance data from Vietnamese provinces, historical demographic data and Earth observation data, such as minimum, maximum and average air temperatures. It also included precipitation, daytime humidity, wind
speed and sea surface temperature. All from the years 2007 to 2016.
‘This is the first instance of such a model being used in dengue epidemiological data to create an early warning system.’
These models were developed to predict the incidence of dengue up to six months ahead. Whilst results showed that the models were more accurate in predicting one to three months ahead, it was not helpful after the fourth month.
However, the ability to predict outbreaks varied across different regions of Vietnam, as well as throughout different times of the year—the summer months from February to June were more accurate, which also coincided with the peak season of
Additionally, results showed a difference between North and South Vietnam provinces. The southern provinces experience the majority of cases during peak season compared to northern provinces; with the exception of Hanoi, the capital city, a
highly urbanised area with 8,700 dengue cases per year over the past decade.
Researchers believe this may be the first early health warning system to be informed by Earth observations, in order to predict dengue epidemics all year long. These models will give a lead time of one to three months, whereas other models
thus far have only provided one or two weeks ahead.
The benefits of a sophisticated forecasting model
Are superensemble models superior to other competing models? It depends, as individual models can be subject to many biases. The superensemble methodology bridges the gap where other models show forecasting inaccuracies and ultimately may
generate more accurate and useful predictions. However, the efficiency of using a superensemble model, in this case, was lost when looking beyond four months ahead.
Moreover, the superensemble modelling has proved useful and reliable in other settings, such as forecasting dengue outbreaks in Puerto Rico and Peru, and influenza outbreaks across the United States.
As the technology improves, we can start understanding more clearly the relationships between disease and climate, as well as the vulnerabilities in society, such as inequality, and how these may exacerbate the impact of disease epidemics. We are now closer than ever to making
science-based decisions to prevent or better prepare for an outbreak, as well as preventing the accompanying losses that come with it.
Forecasting methods still have limitations
This innovative forecasting model is useful but has various limitations. Some of these include not being able to assess the consequences caused by mosquito control interventions, vector (mosquito) population, serotype-specific circulation,
herd immunity and the mobility of people and goods. All these factors are still important in determining the impact of an outbreak.
Moreover, there is still the problem of accounting for underreported cases, the lack of detailed healthcare data at a regional level and confirmed laboratory diagnosis.
‘Researchers believe this is the first early health warning system informed by Earth observations.’
With the current threat of epidemics, it is easy to see the value of investing in better forecasting methods. In the case of the Ebola epidemic in Sierra Leone between 2014 and 2015, real-time forecasting helped make informed decisions, as well as more reliable ones.
This technology has been revolutionary in climate modelling as well as disease modelling. Overall, the study shows promising signs in creating an early warning system for dengue disease outbreaks, which may help guide prevention programmes
such as mosquito control and many public health campaigns.
Featured Image: Vince Gx | Unsplash
Colón-González F.J., Soares Bastos L., Hofmann B. et al. (2021) Probabilistic seasonal dengue forecasting in Vietnam: A modeling study using superensembles. PLoS Med. Volume 18, issue 3,
Funk S., Camacho A., Kucharski J.A. et al. (2019) Assessing the performance of real-time epidemic forecasts: A case study of Ebola in the Western Area region of Sierra Leone, 2014-15. PLOS Computational Biology.
Volume 15, issue 2, e1006785.
Gubler D.J. (2011) Dengue, Urbanization and Globalization: The Unholy Trinity of the 21(st) Century. Tropical medicine and health. Volume 39, issue 4, pages 3-11.
Hii Y.L., Zhu H., Ng N., et al. (2012) Forecast of Dengue Incidence Using Temperature and Rainfall. PLOS Neglected Tropical Diseases.Volume 3, issue 11, e1908.
Johansson M.A, Apfeldorf K.M, Dobson S., et al. (2019) An open challenge to advance probabilistic forecasting for dengue epidemics. PNAS. Volume 116, issue 48, pages 24268-24274.
Naish S., Dale P., Mackenzie J.S. et al. (2014) Climate change and dengue: a critical and systematic review of quantitative modeling approaches. BMC Infectious Diseases. Volume 14, issue 167, pages 14-167.
Rocklöv J. and Tozan Y. (2019) Climate change and the rising infectiousness of dengue. Emerg Top Life Sci. Volume 3, issue 2, pages 133–142.
Schmidt W.P., Suzuki M., Dinh Thiem V., et al. (2011) Population Density, Water Supply, and the Risk of Dengue Fever in Vietnam: Cohort Study and Spatial Analysis.PLoS Med. Volume 8, issue 8, pages
Tompkins A.M., Colón-González F.J., Di Giuseppe F., Namanya D.B. (2019)Dynamical Malaria Forecasts Are Skillful at Regional and Local Scales in Uganda up to 4 Months Ahead. GeoHealth. Volume 3, issue 3,
Tompkins A.M. and Di Giuseppe F. (2015) Potential Predictability of Malaria in Africa Using ECMWF Monthly and Seasonal Climate Forecasts. AMS. Volume 54, pages 521-540.
Yamana T.K., Kandula S. and Shaman J. (2017) Individual versus superensemble forecasts of seasonal influenza outbreaks in the United States. PLOS Computational Biology. Volume 13, issue 11, e1005801.
Yamana T.K., Kandula S. and Shaman J. (2016) Superensemble forecasts of dengue outbreaks. Journal of the Royal Society Interface. Volume 13.
Yun W.T., Stefanova L., Mitra A.K. and Vijaya Kumar T.S.V., Dewar W. and Krishnamurti T.N. (2005) A multi-model superensemble algorithm for seasonal climate prediction using DEMETER forecasts. Tellus A: Dynamic Meteorology and
Oceanography.Volume 7, issue 3, pages 280-289.