2019 is one of the worst years of the dengue epidemic. Scientists and officials are looking for new ways to use data to predict a terminal disease.
In the Philippines, over 622 people died of dengue this year – the majority of whom are children under the age of five.
There are more than 5,100 new cases every week in the country. In at least one province, tents serve as makeshift treatment centers to cope with patient sympathy. The Philippines announced a national epidemic in August.
|(AHC Dengue Image)|
The Philippines are not alone.
In Bangladesh, the dengue epidemic is the worst in the country.
The sheer number of patients with dengue – over 2,000 new cases appear within 24 hours in just 24 hours – creates a huge burden on healthcare systems. Rapid testing kits that could detect early stages of dengue were missing. Demand for blood and platelets is said to outstrip supply in many donation centers.
In Cambodia, representatives from one hospital say the explosion may be the worst they've seen since the hospital opened 20 years ago.
Until now, in 2019, the Angkor Children's Hospital treated over 2,937 patients with dengue – 42% more than at the same time in 2012, which was previously the worst hospital year recorded.
And the cases are more serious than normal.
"Usually, severe cases make up about 10 percent of admissions for dengue," said Dr. Ngeth Pises, medical director of Angkor Hospital on the Asia News Network.
|(AHC Dengue Image)|
"But this year, over 50 percent of cases were hospitalized. We have seen more serious cases than ever before – he added.
The hospital is running out of space for patients and it was decided to spread mattresses on the floors in the hallway.
Not all dengue years are so bad. One of the key features of the disease is that the peak years of the epidemic occur in cycles, usually with an interval of about 3-4 years. But scientists don't quite understand why this is, and more importantly, how to predict when and where a major epidemic will occur.
And this means that even where dengue has long been endemic, places like the Philippines, Bangladesh and Cambodia may be surprised.
This is partly due to the unique ways in which the dengue virus works. But this is also partly due to data collection problems: we need better data and more to predict dengue.
Richard Maude is head of the epidemiology department at the tropical medicine research unit at Mahidol-Oxford. His team from Bangkok provides support to governments in Thailand, Burma and indirectly in the Philippines. Mathematical modeling is one of the forms supported by support.
The team is using new methods to calculate variables that would otherwise be difficult to consider – for example, human movements – to develop models for predicting future dengue outbreaks.
To create a Thai-specific model, people are taken into account, their coming and actions, Maude and his team turned to data creation devices that we all keep in our pockets: our cell phones.
The Maude team took data from Thai phone users that had been cleared of identifying information, transformed this information into a matrix of how many people were moving, and connected it to a model that connects this data with the locations where the dengue time virus appeared.
The real measure of the success of such a model, says Maude, is that "these projections must be good enough for the government to then use them to develop a plan and action based on the projections."
For example, if governments know when the bad year is coming, they may be able to prevent dengue and hospital bed shortages that are currently taking place in Bangladesh, Cambodia and the Philippines.
But there is still a lot of work to be done so that a model like this can be used in this way.
"The performance of these predictive models is good in some locations, but not so good in others, and the result is a lot of different models that work in different areas."
So far, "there is no single predictive model that accurately predicts dengue," said Maude.
"One of the reasons," explained Maude, "is that there are many different factors that affect the likelihood of a dengue outbreak, and these factors are not necessarily well understood."
In addition, complicating the matter, which should logically indicate an increased risk of dengue does not always work in a predictable way.
For example: "We know that you can get a large number of dengue mosquitoes [in this case Aedes aegypti] and then there is no dengue in some places, it does not always happen, "said Maude.
Perhaps the biggest problem holding back these predictive models is data.
"We know about rainfall that causes mosquitoes, we know people spread dengue during travel," said Maude. "But there are also things that we can't take into account – for example, dengue resistance over time in a population."
Data on population resilience is crucial to predict how dengue will circulate in the community, and how serious these cases will be. Patients who recover from a single dengue serotype infection will have lifelong immunity to this particular strand of virus. But subsequent infection with other serotypes increases the risk of developing severe dengue. But, unfortunately, resistance data is simply not collected routinely enough to be included in these models.
Similarly, reliable data on the prevalence of four unique dengue strains or serotypes is difficult for researchers such as Maude. In fact, according to Maude, "in most countries only a minority of people have information about its serotype."
"To date, attempts to develop models using this data are limited," explained Maude.
Climate data – temperature, rainfall, etc. – may be uneven, and data on dengue cases are not always reliable at a very basic level. The number of cases can be overstated – in healthcare systems that decide to treat dengue-like symptoms without testing for dengue – and are understated – in places where people have the virus but are not seeking treatment.
Maude quickly points out that the gaps in dengue data are not necessarily due to a dengue control program in one country.
"The entire healthcare system and population are responsible for this data," says Maude.
To get complete and accurate data, Maude says scientists "rely on a huge number of people to do something consistently." A great question, even in the best conditions.
There is simply no incentive for every doctor to report every case of dengue.
As Maude put it, collecting data is "just more work."
Are there countries already using prognostic models to predict dengue outbreaks?
The only government that Maude is aware of is already using predictive models is Singapore.
One model developed in Singapore with the help of machine learning was able to forecast serious dengue outbreaks in 2013 and 2014 more than 10 weeks in advance, enabling the government to prepare hospital beds, diagnostic kits, deploy additional personnel for mosquito ground control and community coverage.
The city-state partly used information from this predictive model to mobilize information campaigns to raise awareness about dengue when data forecast a bad year.
"This model allows us to definitely warn the public that an epidemic may break out," said Lee-Ching Ng, who is director of the Singapore National Environmental Agency, in a 2016 article on this particular predictive model.
"It's very difficult to be vigilant all the time. You are tired. Public messages cannot be sent all the time, [so] the model suggests when to intensify our message or mobilize the community, "she added.
But Singapore has several key advantages: very high quality data in a very small geographical area.
For other countries, "as well [these models] work depends on what the data looks like in each country, what is available and the format of this information. "
Until the availability and accuracy of data improves, innovation in the area of forecasts will appear slowly.
Quinn Libson (email@example.com) / Asia News Network