Say what? You can predict disease spread with Twitter?


Social media has totally integrated in to our lives. We are constantly updating our status on Facebook or Twitter to let our social network know how we are feeling and what we are doing. For that reason Twitter is an ideal source for data if you want to predict human behavior on a large scale. Researchers from the University of Rochester thought the same way and wanted to make prediction on human behavior on a large scale that would be useful for both individuals and organizations. They therefor decided to try and predict the spread of diseases through Twitter posts. And they succeeded to do so, but how did they do it?

A twitter post contains a couple of simple elements. First of all there are some lines with text in which a person self-reports what he is doing or how he is feeling. From this they were able to detect if a person was having an influenza-type of disease. A twitter post also contains a date and a time, which needless to say helps map out the pattern of disease spread on a more detailed level. However, what they first of all need to map out the pattern of disease spread is a location of someone who is giving of influenza type signals. They achieved to collect this data through the fine-grained GPS location that is attached to a Tweet. By tracking this data from millions of tweets they were able to map out the spread of these influenza type diseases (Couwenberg, 2011; Sadilek, Kautz, & Silenzio, 2012). After that they created an application called GermTracker were you can explore the pattern of the spread of diseases. It also shows you on a map were sick people have been in the last couple of hours and were you currently are through your GPS location and how sick people near you could have impacted your health (Humanaut, 2015; Sadilek, Kautz, & Silenzio, 2012).

GermTracker

However, they took it a step further by combining these patterns of disease spread with dozens of other factors that compose a threat to a persons health like pollution levels.  From this they are able to make predictions eight days in to the future about what your health is going to be like. They do this by combining these different data sets with your GPS locations of the last couple of days. For example, people that take the subway every day, visit bars often or live close to pollution sources are significantly more likely to catch the flu. According to researchers from the University of Rochester these predictions are right 90% of the time. They track around 10 cities over the world which gives them a pretty good idea of what a typical day in terms of diseases look like in these cities based on historical data. The can compare new days with these typical days and issue alerts when they see a rise in the number of sick people in a certain geographical area. On a personal level a person can use this information to make choices that can help him avoid getting sick. For example, you could decide to not take the subway to your work but go by bike. According to the researchers the application could also be of public use by assisting the government in giving of health alerts (Sadilek, Kautz, & Silenzio, 2012).

Personally, I cannot see the benefits of obtaining all this information on a personal level. In my opinion you cannot run from a disease and I wouldn’t want to spend my time on trying to avoid it. But seeing the fact that ten thousands of people use GermTracker on a daily basis, this is apparently a matter of opinion. On a public level however, I think this application has a huge potential when it comes to assisting public health institutions. For example it could help hospitals by alerting them that they can expect a higher number of people coming in with certain diseases and thus help them more effectively deploy human resources. It could also assist hospitals in estimating the number of flu shots that should be available by predicting the chance of a flu epidemic. What do you think about this application? Would you appreciate it that an application lets you know that an hour ago a sick person was at the restaurant where you are now eating a meal? Would you like to know that you are going to be sick in a couple of days or would that be a depressing thought for you? Definitely interesting questions which will probably differ from person to person.

References:

Couwenbergh, H. 2011. ‘De anatomie van een tweet’. Tweetmania [Online], Available: http://twittermania.nl/2011/04/de-anatomie-van-een-tweet/ [19 Oct 2015]

Humanaut. 2015. ‘Germtracker’. Humanaut [Online], Available:
http://humanaut.is/projects/germtracker/ [19 Oct 2015]

Sadilek, A., Kautz, H., & Silenzio, V. 2012. Predicting Disease Transmission from Geo-Tagged Micro-Blog Data. Proceedings of the Twenty-Sixth AAAI Conference on Artificial Intelligence.

Advertisements

One response to “Say what? You can predict disease spread with Twitter?”

  1. hgouiza says :

    I agree on your point that it is useless to run away from a disease. You can not predict whether you are going to be sick or not. It is not worth your time spending on an application as GermTracker. Although doctors and hospitals can be aware about their surrounding and act more quickly to it.

    However, you got my attention on the use of Twitter data. It can be useful. It is becoming more apparent how data mining can be used in grateful matter. Some people are ignoring the fact that a lot of people are complaining on Facebook and Twitter. But when your neighbors tweeted that they were robbed just an hour ago, police can monitor the neighbourhood for your safety.

    By this, people should be more socially active so data miners can do their job. Complaining, about being ill for instance, can be a good thing;)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: