PULSE LAB DIARIES

Using Digital Tools for Disease Surveillance: Mapping Innovations

Sally Jackson
May 20, 2014

Can data innovation provide new insights into rabies and avian influenza outbreaks in Indonesia? This is what we are currently exploring at Pulse Lab Jakarta together with the Food and Agriculture Organization’s Emergency Centre for Transboundary Animal Diseases.

For both avian influenza and rabies, strong surveillance is essential so that there can be an informed and appropriate response (more information on the Indonesian
context can be found at ProMED mail). But increasingly, digital information is being  used to complement data captured through existing surveillance systems. Broadly speaking, new information on disease events could:

• Alert authorities to new outbreaks

• Provide real-time information to improve situational awareness of existing outbreaks

• Identify geographic areas where surveillance may need strengthening

Where to start to identify entry points for innovation? The first step for us has been to conduct a horizon scanning exercise: what new approaches are being used worldwide for disease monitoring that might have not become mainstream yet? Is there anything we can learn from them? What could be translated to the Indonesian context?

In this first blog post, I will summarize the results of this exercise. In future posts I will introduce some of the other work we are embarking on here in Indonesia.
Broadly speaking, existing innovations can be broken down into three different categories:

  • Passively generated information: The data already exists, we just need to use it
  • Actively generated information: Using new methods to get data
  • Information aggregation and visualization

 

Passively Generated Information

This type of information is generated as a result of our day-to-day activities. Information about what people say and do is captured in online media like news,advertising, or social media, and their use of financial, communication, or information 

Since this information is already being produced, it is typically:

  • Low cost
  • High volume/frequency
  • Noisy
  • Representative of technology users (considerations of digital divide essential)

 

Example: GermTracker

GermTracker analyses Tweets using a combination of machine learning and human computation to find meaningful trends about the spread of illnesses. An excerpt from an interview with Adam Sadliek.

“We have always been interested in studying human behavior at scale, and Twitter is a
prime example where you can get enough data to do this. Previously, people have
looked at things like can we predict where people are… you know, location wise based
on what they talk about, and also who they associate with… and that’s interesting. But
we were thinking, can we do even better? Can we infer something that is not obvious
and could be potentially very useful for individuals as well as organizations?”
 

Additional examples:


The Global Public Health Intelligence Network (GPHIN) identifies outbreaks from websites, news wires, local and national newspapers retrieved through news aggregators. It uses machine translation and manual analysis. Social networking sites were used in the surveillance of an avian influenza A
(H7N9) outbreak in 2013. 

Sickweather scans social networks for indicators of illness so you can check for the chance of sickness in your geographic area.

Google uses certain search terms as indicators of dengue and influenza activity. They use aggregated search data to estimate activity around the world in
real-time.

DoctorMe is a mobile app developed to provide health information to users in Thailand. Here is a talk about the opportunity to use it as a participatory surveillance platform.


2. Actively Generated Information

Key characteristic - people actively provide you with information: is designed for purpose

Opportunity cost to those reporting (time they spend away from other activities) so reporters need an incentive (not necessarily financial)

Example; Hawaii State Department of Health’s Foodborne Illness Complaint
app


If you come down with a food-related illness in Hawaii, you only need to travel as the app store to file a complaint. The Illness Reporter App allows you to register information about the establishment or product you are complaining about, and details about the affected person including a comprehensive range of symptoms.

Additional examples:

Citizen reporting

  • Flu Near You and Outbreaks Near Me both use information from HealthMap (see example in next section) to enable the community to report new outbreaks and receive current outbreak information. ‘Flu Near You’ is a website, ‘Outbreaks Near Me’ is an android application.
  • Wildlife Health Event Reporter is a USA-based online tool to gather information from citizen scientists about sick, injured and dead wildlife.
  • Dengue na Web is a tool used in Brazil to enable citizens to contribute information on dengue cases. This data is used for real-time analysis, and to simulate disease spread in the city and intervention scenarios. This was based on the influenza project Gripenet in Spain.
  • Influenzanet is a Europe-wide, web-based flu surveillance system that monitors the activity of influenza-like-illness with the help of citizen volunteers. Here, Daniela Paolotti discusses their experiences implementing this project, including some of the considerations that they made when implementing this system across different countries and cultures.

 


Mobile data collection through surveys:

For mobile data collection through surveys, multiple reporting tools with different
characteristics have been developed and the range can be quite overwhelming. The NOMAD Project is a really valuable source of information for navigating the available options (they currently have 35 different mobile data collection tools listed). Their online selection tool is a quick survey that matches your needs with appropriate solutions.


3. Information Aggregation and Visualization:

No single data source can capture the full picture of the complex environments that we work in. These new data sources should be considered as complementary to
existing sources. Aggregation is important to pull the different sources together, visualization so that the information can be understood and used.

Using Digital Tools for Disease Surveillance: 

An Example: HealthMap

HealthMap “brings together disparate data sources, including online news aggregators, eyewitness reports, expert-curated discussions and validated official reports, to achieve a unified and comprehensive view of the current global state of infectious diseases and their effect on human and animal health”. These disparate data sources include crowdsourcing from ‘Flu Near You’ and ‘Outbreaks Near Me’ (as discussed above).

As this screenshot of HealthMap illustrates that there were two stories about rabies in the past three weeks

Additional examples:

  1. OpenStreetMap is a project to create a free and open map of the entire world, built entirely by volunteers surveying with GPS, digitizing aerial imagery, and collecting and liberating existing sources of geographic data. In response to the recent Ebola outbreak in West Africa, the humanitarian community worked online to map the outbreak area. As of April 15th (20 days in total), 363 contributors had mapped 1.65 million objects, 150,000 buildings, 5100 places, 9,900 landuse polygons and 22,200 highway sections.
  2. EpiSPIDER takes free text from outbreak/epidemic summaries and visualizes it on a map using natural language processing and visualization algorithms. Although the visualization is automated, the data source is ProMED-mail – it is curated manually by experts.
  3. Malaria Atlas Project works to generate new and innovative methods of mapping malaria risk.

What next?

Overall, it can be said that real-time digital information about populations has not been used to its full potential. At Pulse Lab Jakarta, we are working on a proof of concept the use of Twitter for avian flu and rabies monitoring.

In the first run of our ‘taxonomy’ (key words used as selection criteria to extract Tweets), we found 170,000 Tweets about avian influenza, but only 133 on rabies (we
had expected more!). We are now revisiting rabies to see what happened but we suspect that we may find more digital signals by idenitifying more relevant key words.

Do you know of any other useful information sources and tools, or have experiences implementing similar projects? Share them in the comments section. 


Sally Jackson is a Monitoring and Evaluation Specialist at Pulse Lab Jakarta

Headline photo credit: Germtracker by Humanaut

 

 

 

 

Add comment