Turning the public into ‘citizen epidemiologists’

The 2013-2016 West African Ebola virus epidemic that devastated Guinea, Liberia and Sierra Leone, caused small outbreaks in Nigeria, Mali and Senegal, and later imported into the UK, US, and Spain, was an important wake-up call to public health agencies around the world. Infectious diseases such as Ebola have traditionally been considered to be isolated events in developing countries – but the recent outbreak illustrated that this is no longer the case in an increasingly globalized and interconnected world. The outbreak was certainly a manifestation of West Africa’s weak health systems, but more importantly, highlighted the poor communication infrastructure in place that prohibited quick and timely information flow, as the disease spread quickly from a small village in Guinea to the rest of the world.

Existing public health reporting in developing countries is primarily a linear process,[1] as depicted in the figure below (A). It is a manual, paper-based data and information collection process, with poor geographic coverage and huge time lags in communication. However, increasing digitization and mobile phone penetration, particularly in Sub-Saharan Africa, provide a huge opportunity for more direct communication between groups, enabling more timely information flow and faster detection of emerging diseases (B).

Figure 1[2]

HealthMap has leveraged this opportunity to develop a free online disease mapping tool that provides a comprehensive global view of the current state of infectious diseases. Every hour, 24 hours a day, HealthMap automatically acquires information from a variety of freely accessible websites across 15 different languages, and then filters and organizes the data based on a machine-learning algorithm. Since its inception in 2006, the algorithm continues to be improved based on feedback from analysts, and sifts through the noise by cross-validating the acquired information and weighting by news source reliability (e.g. higher weight to multilateral agencies) and by the number of unique data sources (e.g. higher weight to multiple sources).[3] It is used primarily for early disease detection but also to raise awareness among local health departments, public health agencies, and international travelers to high-risk countries. HealthMap’s most recent claim to fame was its tracking of early press and social media mentions of a hemorrhagic fever in West Africa, a week before the WHO officially identified it as Ebola.

Figure 2[4]

HealthMap has extraordinary potential to provide real-time and highly local information, and detect emerging outbreaks, much before governments and multilateral agencies do. However, there are a number of challenges it will need to address to further add value to global disease surveillance efforts.

Leveraging information from ‘dumb’ phones not connected to global information network

High mobile penetration, greater access to mobile broadband, significant infrastructure investment in 3G/4G networks – these are all examples of Africa’s rapid digitization[5] and highlights the increasing sources of information that HealthMap can tap into. While Africa’s digital transformation is attracting much global attention, it is important to remember that for the 80% that do own a phone, 82% only have access to basic (or feature) phones[6] and that the majority of households are still isolated from the digital world. Such underdeveloped regions pose the greatest risk of disease outbreaks given their poor infrastructure, and thus investment to connect these communities to global information networks is crucial. Fonetwish is one such example, which uses the USSD[7] protocol to connect users to social media networks via a text-only functionality.[8]

Distinguishing signal from noise via more open collaboration

Google Flu Trends (GFT) provided estimates of influenza activity by analyzing trends in Google search queries, such as headache and chills. Such search terms were then correlated with flu outbreak data from the CDC. GFT closed its doors after failing to predict the peak of the 2013 flu season by 140 percent[9]^,[10], highlighting the pitfalls of big data and letting online algorithms make critical health decisions. While GFT algorithm flaws have since been identified, e.g. not taking into account changes in search behavior over time, the opaque nature of GFT’s data and prediction methodology prevented a meaningful feedback loop with its users. Had GFT been more transparent and allowed for crowdsourcing of ideas, it may have been able to more quickly distinguish ‘good data’ from ‘bad data’. HealthMap should take these lessons into account as it acquires and maps its disease data.

Figure 3[11]

Hedging against risk of state censorship

Some governments may censor disease outbreak information to the public due to fear of negative coverage, such as China during the 2003 SARS outbreak or Saudi Arabia during the 2012-2013 MERS outbreak.[12] State censorship, particularly in the early crucial weeks, impedes information flow and gives the disease time to spread and potentially mutate, increasing the risk of a widespread outbreak. To hedge against such risks, HealthMap will need to work closely with local partners, multilateral agencies and NGOs in country to obtain information that otherwise may be lost under government censorship.

(795 words)

[1] CHUNARA R, FREIFELD CC, BROWNSTEIN JS. New technologies for reporting real-time emergent infections. Parasitology. 2012;139(14):1843-1851. doi:10.1017/S0031182012000923.

[2] Ibid.

[3] Brownstein JS, Freifeld CC, Reis BY, Mandl KD (2008) Surveillance Sans Frontières: Internet-Based Emerging Infectious Disease Intelligence and the HealthMap Project. PLoS Med 5(7): e151. doi:10.1371/journal.pmed.0050151

[4] Screenshot of HealthMap.

[5] McKinsey Global Institute, “Lions go digital: The Internet’s transformative potential in Africa,” http://www.mckinsey.com/industries/high-tech/our-insights/lions-go-digital-the-internets-transformative-potential-in-africa, accessed November 2016.

[6] Consumer Technology Association, “How Mobile Phones Are Changing the Developing World,” July 27, 2015, https://www.cta.tech/News/Blog/Articles/2015/July/How-Mobile-Phones-Are-Changing-the-Developing-Worl.aspx, accessed November 2016.

[7] USSD: Unstructured Supplementary Service Data

[8] Quartz India, “Two-thirds of the world’s mobiles are dumb phones. Meet the company getting them online,” June 8, 2014, http://qz.com/217909/two-thirds-of-the-worlds-mobiles-are-dumb-phones-meet-the-company-getting-them-online/, accessed November 2016.

[9] Wired, “What We Can Learn From the Epic Failure of Google Flu Trends,” October 1, 2015, https://www.wired.com/2015/10/can-learn-epic-failure-google-flu-trends/all/1, accessed November 2016.

[10] Nature, “When Google got flu wrong,” February 13, 2013, http://www.nature.com/news/when-google-got-flu-wrong-1.12413, accessed November 2016.

[11] Ibid.

[12] Wired, “Censorship Doesn’t Just Stifle Speech — It Can Spread Disease,” August 21, 2013, https://www.wired.com/2013/08/ap_mers/all/1, accessed November 2016.