New Zealand’s health minister, Jonathan Coleman, has launched a project that will seek to establish if tracking trends on social media and using ‘unconventional data’ can help predict outbreaks and further improve responses to epidemics.
Similar approaches have already been demonstrated overseas, and reported in a number of academic papers.
Dr Coleman said: “People often talk about being unwell on social media, so trends can be detected on platforms like Facebook and Twitter. Picking up on trends could help us to put the appropriate measures in place earlier to prevent disease spread, and ensure sufficient stocks of medicines are available.”
He added: “The Ministry is also harnessing a wide range of data for this project, such as anonymised information about school absenteeism, employee sick leave, pharmacy sales of over-the-counter medicines, Healthline calls and tissue sales.
“Claims that luxury soft tissue sales surge at the start of influenza outbreaks are also being analysed to see whether not just the sale volumes but the types of products can act as an early epidemic warning.”
“This project builds on our existing monitoring programmes which work well to identify trends in communicable diseases using traditional methods, such as surveillance of lab results and data from general practices.”
The Department of Health is conducting an online survey that asks people if they have ever posted information social media about themselves or their family’s illnesses.
The Department said the project was being funded through Vote Health, which this year received $888m in funding.
The power of social connections
Back in 2010 in a research paper Social Network Sensors for Early Detection of Contagious Outbreaks by Nicholas A Christakis and James H Fowler reported how they had studied a flu outbreak at Harvard College in late 2009 in an attempt to better predict the likelihood of an epidemic.
The core of their method was that “individuals near the centre of a social network are likely to be infected sooner during the course of an outbreak, on average, than those at the periphery.”
They said: “We followed 744 students who were either members of a group of randomly chosen individuals or a group of their friends. Based on clinical diagnoses, the progression of the epidemic in the friend group occurred 13.9 days in advance of the randomly chosen group (ie, the population as a whole).
They found that the friend group also showed a significant lead time on day 16 of the epidemic, a full 46 days before the peak in daily incidence in the population as a whole.
They concluded: “This sensor method could provide significant additional time to react to epidemics in small or large populations under surveillance,” adding: “The method could in principle be generalised to other biological, psychological, informational, or behavioural contagions that spread in networks.”
The same principle has been applied to analyse Twitter feeds around Hurricane Sandy, which struck the East Coast of the US in 2012.
Researchers found “differences in users’ network centrality [how well they are connected through social media to other people] effectively translate into moderate awareness advantage (up to 26 hours).”
- Australia’s war on maths blessed with gong at Pwnie Awards
- Bureau of Meteorology to offer access to data via APIs
- Qrious launches big data powered visitor movement service
- MercyAscot deploys Qlik analytics to boost efficiencies
- Startup Floodlight Analytics helps venues make sense of data
- Guidelines aim to promote responsible, trusted data sharing