Could the internet help prevent international health emergencies?

In November 2015, a small team that scans and analyses online content for the early detection of infectious disease outbreaks noted a marked increase in reports of skin rashes in Brazil. However, they had no formal channel to notify the Brazilian health authorities or the World Health Organization. A few weeks later, the outbreak of Zika virus was announced to the world; skin rash is one of the most common symptoms of infection with Zika virus.

The provision of online data for the early indication and tracking of disease outbreaks or other health events is known as digital disease detection (DDD). It includes a wide range of methods, from the voluntary reporting of symptoms by individuals to the scanning of media sources and the analysis of posts on social media, mobile phone data to map population movement and patterns of keywords typed in search engines over time. These approaches have already been used to identify and investigate outbreaks of infectious diseases such as influenza, Ebola and Zika as well as chronic health conditions like insomnia and obesity. They have also been used to model disease emergence hotspots and to gather information on other public health issues such as gun violence or quality of health care.

Although DDD has been used since the mid-1990s, it was not until the mid-2000s that its profile really increased when Google developed its Flu Trends − promoted as detecting influenza outbreaks ahead of official public health surveillance systems. While DDD generated considerable enthusiasm in the public health community, three major questions arose. First, how could real health events be distinguished from background noise? Second, what was its true predictive ability? Third, and most importantly, what was the added public health value?

These questions were highlighted in 2012-13 when Flu Trends predicted twice the number of influenza cases than actually occurred. In 2015, Google stopped making Flu Trends estimates publically available. However, this setback does not negate the value of DDD. It simply emphasizes the risks of relying solely on a new technology to monitor and predict disease patterns instead of integrating it as one tool to complement others that have a demonstrable track record.

As this nascent field begins to mature, there are improvements in the validity, accuracy and utility of the data it generates. It is increasingly being used by the public health community to investigate health issues. For example, mobile phone data were used to track population movement during the West African Ebola outbreak in order to predict where new cases may arise, and Twitter feeds have been used to identify and contact individuals affected by outbreaks of foodborne disease.

As the technology develops, including in the area of ‘big data’, opportunities to use DDD to anticipate, detect and monitor health issues such as outbreaks will increase, and the international community needs to be ready to capitalize on them. To do so, there are new questions to be answered.

What is the status of the information derived from DDD? Most of the organizations generating these data are in the academic, private, or not-for-profit sector, and sit outside the normal government-owned disease surveillance and response system. There are few examples of official public health authorities routinely using and acting upon DDD data. Unfortunately, this can lead to situations where a health event is detected, but the responsibility for acting upon the information is not clear. The organizations generating the data often do not have, nor can they be expected to have, any response capacity.

How should DDD be incorporated into the broader, formal surveillance landscape, and how can it be systematically tied to a verification and response mechanism? The public health community’s increasing use of online data raises important ethical and legal issues. For example, how to deal with data that can be considered private and that are most commonly not collected for health purposes. There is currently little understanding of how the general public might perceive the use of their online data for public health purposes, and it remains to be determined what might constitute a reasonable balance between individual rights and the common good. There is also a lack of good practice evidence or standards applicable to DDD.

The technologies will soon have developed to the point where they will be able to provide reliable and robust public health information to complement more traditional ‘formal’ surveillance systems on a routine basis. The place of DDD within the broader landscape and the technical, ethical and legal issues raised has to be tackled now to ensure the potential is realized. It will be vital to engage national governments and public health agencies to gain the necessary traction and ownership to build sustainable systems and to future-proof them as far as possible. Not doing so may lead to critical weaknesses in public health systems, not least in the ability to detect and respond to the next Ebola or Zika.