Twitter feeds and online news were faster than
traditional surveillance in detecting the onset and spread of the cholera
epidemic after the January 2010 earthquake in Haiti, according to study results
published in the American Journal of Tropical Medicine and Hygiene.
To determine whether trends in the volume of online
social and news media correlated with officially reported disease measures,
researchers examined data from the first 100 days of the Haitian cholera
Researchers accumulated data from three sources: reports
from the Haitian Ministry of Public Health (MSPP) that consisted of tallies of
cholera cases, hospitalizations and deaths, often published in time-delayed
batches; Twitter, a public social media service examined for any French and
English mention of the word “cholera;” and HealthMap, an automated
surveillance platform that constantly identifies, characterizes and maps
outbreaks and epidemic events.
Incoming data were organized into three time periods to
correlate with the initial phase of outbreak (Oct. 20-Nov. 3, 2010); the
increase in cases during Hurricane Tomas (Nov. 3-Dec. 1, 2010); and the first
100 days of the outbreak (Oct. 20, 2010-Jan. 28, 2011).
According to the study, data compiled from HealthMap and
Twitter demonstrated the highest correlation in time with 0-day lags for all
three time periods (Oct. 20-Nov. 3: 95% CI, 0.64-0.91; Nov. 3-Dec. 1: 95% CI,
0.64-0.86; and Oct. 20-Jan. 28: 95% CI, 0.75-0.84).
Researchers determined that although data collected from
HealthMap and Twitter correlated best with MSPP data provided with a 1-day lag,
the availability of real-time information (as opposed to MSPP 2-week release
delay) allowed earlier estimates to be compiled from these data sources than
from traditional surveillance sources.
When comparing MSPP hospitalization data with HealthMap
data, an improved correlation was observed in the first two time periods (Oct.
20-Nov. 3: 95% CI, 0.55-0.88; Nov. 3-Dec. 1: 95% CI, 0.63-0.85), with poorer
correlation in the third time period (95% CI, 0.29-0.51). Comparisons between
MSPP data and Twitter data demonstrated similarly enhanced correlation in the
first two periods (Oct. 20-Nov. 3: 95% CI, 0.71-0.93; Nov. 3-Dec. 1: 95% CI,
0.36-0.72), but less correlation in the third period (95% CI, 0.13-0.37).
“When we analyzed news and Twitter feeds from the
early days of the epidemic in 2010, we found they could be mined for valuable
information on the cholera outbreak that was available up to 2 weeks ahead of
surveillance reports issued by the government health ministry,” study
researcher Rumi Chunara, PhD, said in a press release. “The
techniques we employed eventually could be used around the world as an
affordable and efficient way to quickly detect the onset of an epidemic and
then intervene with such things as vaccines and antibiotics.”
Disclosure: The researchers report financial
support from Google.org, the National Library of Medicine, NIH and the National
Institute of Allergy and Infectious Diseases.
Andi L. Shane, MD MPH, MSc
This retrospective study compared information gathered from the first 100 days of the cholera outbreak in Haiti from Twitter, HealthMap and the Haitian Ministry of Public Health (MSPP). Reports from the social media sources, HealthMap and Twitter were disseminated more rapidly than those from governmental sources. Good correlation between numbers of cases over time was noted between the social media and official sources.
As pointed out by the authors from Harvard and McGill Universities, there is utility to the use of social media as a reporting mechanism for infectious disease outbreaks, especially in the early stages when a cluster of infections may not be recognized as an outbreak. Social media may also be instrumental during the initial stages of an outbreak when formal surveillance systems are not established. The real-time availability of information from social media facilitates a rapid understanding of the reproductive-number that may be rapidly translated into response.
Providers in Emergency Department and primary clinic settings are frequently the first to observe an unusual manifestation of an infection or a notable cluster of infections. Traditional routes of reporting to local and state health departments may result in delays in release of information to other providers and the public, as involved persons are interviewed and cases confirmed. Social media has revolutionized outbreak surveillance as it has facilitated reporting by anyone with access to a number of modalities. Facebook was instrumental in understanding the epidemiology of an influenza-like illness that affected more than 80 international attendees at a conference in Los Angeles in February of 2011. An initial posting on Facebook resulted in responses from cases around the world that was utilized by the Centers for Disease Control and Prevention to disseminate and online questionnaire and information about testing and management.
Social media is extremely beneficial in obtaining real-time information about disease clusters and outbreaks as demonstrated by the comparative data from the Haitian cholera outbreak. However, the importance of verification and the dissemination of accurate information while emphasizing awareness and reducing fear, remains a challenge.
– Andi L. Shane, MD MPH, MSc
Assistant professor, division of infectious disease, department of pediatrics
Emory University School of Medicine
Disclosure: Dr. Shane reports no relevant financial disclosures.