Facebook: An Innovative Influenza Pandemic Early Warning System

Facebook: An Innovative Influenza Pandemic Early Warning System

Presented by Chen Luo (MBCHB 4) Supervisors: Dr Chris Paton and Dr Robyn Whittaker

Background

• No introduction needed. Influenza is a serious disease – 250,000 – 500 000 deaths worldwide/yr among the

vulnerable (WHO 2003)

– Lost productivity: $71-167 billion (USA)– Here to stay. Rapidly evolving virus – swine flu

• Current surveillance – clinical and virology based on network of primary health doctors

• Disadvantages: 1-2 week reporting lag and health infrastructure

• Facebook 150 million worldwide user base

Literature Review

• Databases searched: MEDLINE, EMBASE, SCOPUS, ISI Web of Knowledge, INSPEC

• Latest news: Google blog search and Google• Hand searching of all relevant articles• 2 prong standardized search strategy used

– 1) Social Networks AND influenza– 2) Web Surveillance AND influenza

• 305 articles– 1) 0 articles– 2) 6 articles

MEDLINE Search

• 1 (OSN or (social network* adj2 (site* or service or provider* or website* or online)) or facebook or myspace or orkut or bebo or friendster or hi5) (295)

• 2 Disease Outbreaks/ (50100)• 3 Influenza, Human/ (20693)• 4 (flu or ILI or influenza like illness or influenza-like-illness or influenza-like

illness) (7406)• 5 2 or 3 or 4 (71592)• 6 1 and 5 (0)• 7 ((Web or internet or net or online) adj3 (Data Collection or survey or

questionnaire* or research or feedback or surveillance or experiment)) (3519)

• 8 5 and 7 (43)

Different Approaches

• 1) Syndromic Surveillance: indirect measure of influenza– Google Flu Trends: health seeking search terms– Global Public Health Intelligence Network: automated algorithm for

news reports

• 2) Primary Doctor (GP) web-based reporting – REALFLU sentinel GP electronic ILI report

• 3) Web-based self-reporting– Pioneer: Great Influenza Survey (GIS) – Netherlands

• 2003-2004 Flu season, 13 300 active participants• Initial demographic questionnaire• Weekly ILI symptoms questionnaire via email• Average correlation: GIS vs Sentinel GP

Gripenet

• Superseded GIS – 2006-2007 Flu season– Netherlands (19,623), Belgium (7,025), Portugal (3118) participants

Influenza Tracking Project

• Inspired by Citizen Science movement• Pilot study – summer studentship• Improve on previous research by

– 1) Simplification: lay person, easy to use– 2) Using social networks to increase participation– 3) Collecting location (IP data) to track influenza globally

• Similar model used: initial demographic questionnaire and weekly symptom questionnaire but a twist – Facebook application

1. Simplification

• Problem: 13 ILI symptoms asked in GIS/Gripenet– Runny nose? Sore throat? Muscle pain? Abdominal pain? Nausea? Sudden Fever (No/Yes/Don’t

Know), How high was the fever? Cough? Headache? Chest pain? Diarrhea? Cold Shivers? Irritated Eyes?

• Complicated case definition; by “experts”– Definition of ILI: Fever > 38C started suddenly plus headache or muscle pain, plus at least one

respiratory symptom

• No current consensus on case definitions• Solution: Thursky (2003): Symptom Triad – Fever (subjective),

cough and fatigue – Effective: more sensitive and specific than CDC. Sensitivity (43.5-

75.1%), specificity (46.6-80.3%) and PPV (23.3-59.7%). – Simple

2. Organic Growth

• GIS/Gripenet used: TV, radio interviews, information posted to Universities/schools and even competitions (!)

• Participant growth dependant on expensive media campaign

• Student = poor• But! Facebook applications engineered to spread.. Like..

(Annoying) Fortune cookie

• Programmed application to be “viral”• Via “newsfeeds”• “Invitations”• “Profile badges”

• Demonstrated later

3. IP Data

• Information which tells you city/country where internet accessed

• Worldwide real time surveillance

How we build the program

• Programming was relatively straight forward– API (Application Program Interface): access to Facebook functions

with just a line of code– Official resources like the Developer Centre provides

documentation and examples– Open source: PHP and MYSQLI – (Inelegantly) program the application based on freely availiable

PHP or MYSQL tutorials on the net– Design is based on copying and pasting Facebook CSS

• Google = best friend

The website

• Website for explanation• www.influenzatracking.com

• Mass media exposure to get “critical mass” for viral spread like a nuclear reaction

IP address automatically stored

Invitations for “viral” spread

Profile Badges for “viral” spread

Newsfeeds (like a newspaper/stalker report of what your friends in your social network are doing) for “viral” spread

Weekly email notification

• Referred by mass media/blogs

Summary Slide

www.influenzatracking.comMedia

Exposure

Sign up

Flu Symptoms Questionnaire

Weekly email reminders

“Viral” Spread mechanisms

Beta Results

• Beta-testing started in 28/06/09• Beta testers - 73 participants• 52 participants filled more than 1 questionnaire• 71% conversion rate• Demographics:

– Age: 30 yrs old (mean)– Gender: 37 M 36 F– Education: 83% of users had tertiary or postgraduate degree– Vaccination: 37% vaccination rate

• 26 application “fans”

Beta Results

• Median days between questionnaire: 8.0• Future calculation:

– No of ILI/no. of total participants in a city/country– World map

• This presentation marks the end of beta-testing and the start of the PR drive public

Plans for the Future

• Mass PR drive• Please join! www.influenzatracking.com• Please do blog/spread the word!

• Research partnerships:– Epidemiologist– Biostatistician– If you can help in any way– [email protected]

Thank you for listening!

• Any questions?