Consider the following hypothetical scenario: Working together, a public health authority, a health management organization and some major food processors have set up a non-profit food safety monitoring center. This center has created a website that invites members of the public to report possible food-related illnesses, and has widely distributed a free smartphone app for the same purpose. Between 8 and 9 pm one evening, the center receives four times the usual number of illness reports. Automated summary statistics show that most of them originate in one city, and are associated with eating at a specific fast food chain. The symptoms appear to indicate an E. coli outbreak. The monitoring center then issues an alert to hospitals, clinics and other medical care providers in the region, requesting prompt reports of any suspected cases of food poisoning. Later that night, consumers in two nearby states report similar symptoms, and more than half of them report eating at the same fast food chain. As the night progresses, the center receives several reports of food-related illnesses from hospitals and clinics, including three positive diagnoses of E. coli 0157:H7.

One hundred plus reports within a few hours is alarming! So the center begins to monitor Twitter for posts about food poisoning or E. coli, and observes a spike in the number of tweets about sickness after eating burgers, originating in the same three states. The next morning, the center uses the online diagnosis application to “push” a question to all recent users in the area asking them to report any illnesses that may relate to eating fast foods. The center quickly receives a high volume of responses about sickness after eating at the same chain. That afternoon, the center notifies the fast food chain and the FDA, and issues an alert to public health officials in all three states on Monday afternoon. The fast food chain initiates a voluntary recall, and public health officials issue a public warning. The outbreak is quickly controlled, and illnesses and deaths are minimized.

Although this is a fictitious scenario, it is entirely possible from a technological point of view. The question is, is such a system feasible from a practical point of view, and how much benefit could it provide?

Reality Check Compare this scenario to existing public health processes. How long would it take for the outbreak to be recognized and controlled? Several hours usually pass between the time when a person notices symptoms of illness and the time when the condition becomes serious enough to consult a health care professional. After the person sees a doctor, tests may be needed, and it may take 24 hours or more for the results to return. The tests may or may not indicate the cause of the illness. At this point, the patient’s symptoms may get worse while the doctor forms a diagnosis and attempts to stabilize the condition. If the cause is recognized as a disease for which there is a mandatory reporting protocol, the case will then be reported to public health authorities. Once public health officials receive notice of a possible foodborne disease, they begin by interviewing the patients and their families to find a common cause. If a source is pinpointed, the public health department visits the location or takes a sample of the food and runs tests to determine whether this is the probable cause of illness. If a specific food is confirmed as the source, a recall will be issued. A public statement will then be made to inform the public about the potential danger of the particular tainted food item. How long does this process take? Weeks or even months may pass before a recall actually happens, by which time many more people can contract the infection. The 2010 outbreak of Salmonella linked to eggs from Wright County Egg, which sickened almost 2,000 people provides a good example of this. The first report of illness was received in July, but the product recall did not happen until August 13, 2010. Another outbreak of Salmonella, associated with ground turkey produced by Cargill, affected 136 people in 34 states. The first illness was reported on February 27, but the product was not recalled until August 3 – just over five months later (Marler, 2012). By the time a recall takes place, a contaminated product has often been distributed, divided, repackaged and resold. Consequently, it is often necessary to recall a large quantity of food across a large geographical area, rather than withdrawing only the specific shipments or batches where the problem originated. This takes still more time to implement. Furthermore, it increases the cost to suppliers. In 2011, the Microbiological Data Program, a produce surveillance initiative housed in the USDA that was cancelled at the beginning of 2013 by budget cuts, identified cantaloupe in New York that were contaminated with Listeria. Unsanitary conditions were found at the supplier’s packing shed, and a recall was initiated on July 28 and expanded two days later. The Food and Drug Administration (FDA) warned the public not to eat the recalled cantaloupe. Ultimately, the supplier recalled an entire growing season’s worth of cantaloupes and honeydew melons in 18 states. Nevertheless, 147 people were sickened by the contaminated cantaloupes and 33 died as a result of their infections. The total cost of the outbreak to the victims, the industry and government has been estimated at $100 million. A New Model: Consumer and Private Sector Participation Using Social Media This article began by describing a hypothetical scenario in which the government and industry cooperatively build a monitoring center, which uses social media to enable effective information exchange and cooperation among public authorities, consumers, and the private sector. As noted already, there are no overwhelming technological obstacles that would prevent the implementation of such a system. In some jurisdictions, parts of such a system already exist. The Hawai’i Department of Health, for example, has created a free iPhone app called “Foodborne Illness Reporter.” It enables iPhone users who suspect that they have contracted a foodborne illness to report the establishment (restaurant or food seller), the suspected cause, the people affected, and other information. Information submitted using the app is confidential, and is shared with the U.S. Centers for Disease Control and Prevention. (see AppShopper, category “Medical”) A properly designed system that uses social media effectively would deliver significant benefits:

–       Drastically reduced response time: By recognizing unusual patterns in data that have been self-generated by consumers, the system can recognize possible food contamination outbreaks even before any of the patients or doctors have reason to suspect such an event. The above scenario used a self-diagnosis application as the data source, but many other possibilities exist. For example, Google is developing a capability called Google Insight, which can provide statistical analyses of searches for particular terms. At a conference which this author attended, the Chief Economist of Google showed a one-day lag between a peak in searches for ‘vodka’ and a peak in searches for ‘hangover’. Clearly, public health information could be obtained by looking for peaks in searches for terms related to illnesses.

Online forums and chat rooms can be another valuable source. For example, Mediator, a diabetes drug launched in France in 1976, was suspected of being unsafe. It was used “off label” for weight loss, and appeared to play a role in several deaths. By 2003, online forums were hosting hundreds of patient conversations about Mediator, but simple word scanning did not reveal any clear pattern. In 2006, however, the forums were analyzed using “argument mining.” This technique revealed a heavy focus on problems related to the drug. Mediator was finally withdrawn from the French market in 2009 (Buahin et al., 2012).

–       The cost of monitoring and reacting can be reduced: As already noted, the MDP program will not exist in 2013 due to budget cuts. Web applications such as self-diagnosis tools (e.g. and search engines are self-funded: They make money for the companies that build them, either by generating advertising and “click through” revenues, or by making potential customers aware of services and products that the builder of the web site offers. By working with the builders of these web sites to give these applications the ability to recognize unusual patterns that may indicate health crises, public health authorities could create wide-ranging and effective monitoring programs at very little cost.

–       Industry cooperation can be increased: Using social media would give both public health experts and industry participants much more data about any event much more quickly. This would make it easier to recognize a problem while it is still small in scale, and to identify the correct scope of any recall more precisely. Thus, the cost to a supplier may be considerably reduced, because a smaller recall may be sufficient to prevent further spread of the pathogen.

Companies in the food processing and distribution industry can maximize the benefits they would receive from such a system in many ways. They can select employees who are skilled at collecting and integrating varying types of information in order to recognize outbreaks more quickly. A real-time, interactive social system can help to identify the most effective actions (e.g., monitoring, diagnosis, treatment, reporting, and/or product recall) that the company could take, thereby minimizing harm to customers and possible liability. (Buahin et al., 2012)

–       If a large outbreak occurs, consumers can participate actively in collecting information and responding to the crisis: An automated crisis mapping platform called Ushahidi (Swahili for “witness”) is already available for free, and could easily be adapted for public health use. This platform uses free “open source” software to issue alerts, receive voluntary reports from any users within a specific radius pertaining to those alerts, and then summarize and analyze those reports. The UN, World Bank and many NGOs have used Ushahidi to help address and track such emergencies as food and medicine shortages across Africa, violent incidents in Gaza, election-related problems in Mexico and India, the revolution in Syria, and swine flu.

As Tapscott (2012) suggests, it has become “increasingly clear that we can rethink and rebuild many industries and sectors of society on a profoundly new, open, networked model” (p. 8). The model described above applies this very general concept to a specific government activity: ensuring the safety of the food supply. Although Tapscott writes primarily in buzzwords, the practical concept behind those words is valid, and offers a significant opportunity. Social media offer a way of collecting large quantities of data items which mean very little individually, but which can reveal significant and actionable patterns when integrated and analyzed. This in turn enables consumers to make a valuable contribution without any need to be experts in food processing or medical diagnosis, and provides much better information to the public and private sectors alike. In effect, consumers, suppliers and regulators become partners that self-organize, contribute to the total sum of knowledge, share information, support each other and jointly manage health and safety. It is particularly significant to note that many elements of a food safety reporting system based on social media have already been created, or are likely to be created, without requiring any government funding or subsidization. This is strong evidence that providing such services as these may be cost-free after tax deductions are considered, or profitable enough that the private sector will take on a significant part of the work. Google clearly believes that Google Insight will earn its keep, and several surprisingly effective self-diagnosis tools (presumably supported by advertising) are already easily accessible. Another example arises from the upcoming shutdown of the Microbiological Data Program. Bill Marler, a leading American food safety lawyer, is considering funding a similar program of his own. Rather than depend on the government to collect data and initiate recalls, Marler plans to collect such information and make it publicly available to the government, consumers and producers. Social media offer a growing potential for structural change and greater efficiency in many sectors, and food safety is no exception. As government budgets continually decrease, and food safety agencies struggle to recognize and prevent food borne illnesses, social media technology can play a key role in improving the effectiveness and speed of response while minimizing costs.


Marler, B. (2012, July 22). Publishers platform: How can we stop outbreaks earlier. Food Safety News. Retrieved from: Marler, B. (2012, Aug 12). Publisher’s platform: Ode to the microbiological data program. Food Safety News. Retrieved from: Tapscott, D. (2011) Macrowikinomics: Rebooting Business and the World. Toronto, ON: Portfolio Penguin Canada. Harrysson, M., Metayer, E., and Sarrazin, H., How ‘social intelligence’ can guide decisions. Mckinsey Quarterly. Retrieved from: Buahin, J., Chui, M., and Manyika, J. (2012, November) Capturing business value with social technologies. Mckinsey Quarterly. Retrievved from: Acheson, D.W.K. (2012, July 12) Should the microbiological data program be saved? Maybe … maybe not… Food Safety News. Retrieved from: “Illness Reporter” in AppShopper, category “Medical” <>

  • Great article.  I would love to see something like that put in place.  I think that a great deal could be accomplished even without apps.  For example, simply monitoring google searches could identify an unusual number of people searching for “bloody diarrhea”

  • Great article.  I would love to see something like this implemented.  I think that a great deal could be accomplished without a special app though.  For example, monitoring Google searches could reveal an unusual number of people in California searching for “bloody diarrhea”.

    • Thank you Kristin! Hopefully it won’t be too long until we see it in action. Any suggestions on how to make this proposal better would be highly appreciated. 

  • Seth

    Great article, but it looks like a site/service something like the one being described in the article has already been created by a company called Food Sentry.

  • husna

     The hypothetical scenario can turn into a brilliant new business proposal, if funding is obtained by private companies:)!

      The iphone App created by the Hawaii’s department of Health can be a working model for the rest of the states for all the significant reasons the author has listed. It might also reduce the workload of individuals that “burn the candles at both ends” from the farm to the fork, and unfortunately at the hospital as well.

    • Thank you husna for your encouraging comments. I agree, this idea can be implemented by funding from private entities. The Ushahidi platform was utilized to help with disaster at extremely low coast and even in a basement in Boston. This proposal can be taken up by anyone but hopefully to the benefit of the public. 

  • Seth

    Not sure why my comment was deleted, but again, thanks Adam for a good article, and I just wanted to share that a service similar to the one described above has been created, and is called Food Sentry.


    Interesting article – the scheme described above may have the potential to shorten the reponse time once an outbreak has occured.  Shortening the response time may reduce both the spread of the outbreak and the extent of any recall.  The described approach however remains reactive, while the goal of the food industry is to use preventative measures to stop contaminated products from reaching the consumer and causing outbreaks.

    One drawback that is not mentioned in the article is the different incubation periods that can occur between when contaminated food is consumed and the onset of symptoms.  For Salmonella this can be 1-3 days, E. coli 3-4 days and for Listeria the one average I have seen is around 21 days and it can be much longer.  Considering the scenario described in the article where people started experiencing and posting about symptoms starting at 8-9 pm at night, are they going to relate this back to what or where they ate lunch or dinner that day generating a false trail where as the actual source is where they ate one or more days earlier.


  • Traceability has come a long way in the last few years, especially in the produce industry.  That, coupled with something like described in the article would go a long way to really shortening the amount of time to track down contaminated food.  Finding the commonality among people getting sick is still difficult and this would be a good way to pin down potential sources much quicker

  • I like the concept.  Couple with better traceability of food products these days,  a data base of food born illnesses would do a lot to shorten the length of time it takes to track down contaminated food.  If something like this would have been in place a couple summers ago, perhaps some lives could have been saved.

  • I’ve been in the tech industry for over 25 years. It can be easy to fall into the trap of seeing technology, especially new technology, as the savior for every problem. Sometimes it is, and sometimes it isn’t. I believe your social media system is a case of the latter. 

    Look at the state of social media. It is, at best, chaotic. That can make it exciting, but not necessarily accurate. 
    During an event such as Sandy, you can see first person reports faster than anything that can be reported on the media. Very exciting, and seemingly informative. However, unless you know the people involved, you have no way of knowing how accurate, or truthful, the communications are. Information is good, but no information is usually better than wrong information, or deliberate misinformation. You don’t show many human checks for reliability in your system. Yet you want to issue alerts to hospitals and others based solely on non-validated, totally spontaneous and unverified data. How long before said system loses all credibility for ‘alerts’ that are nothing more than a bunch of high school kids gaming your system?You also don’t factor in human suggestability. If one person coughs in a room, in five minutes, a dozen will be coughing. Give it time, most of the room will be coughing. One person tweets they ate at Madame Joe’s, and they don’t feel so good. Someone suggests they submit a report to your system. A dozen people read the tweet and think to themselves, they don’t feel so great, either. They report to your system, “just in case”. Chances are the people are geographically distributed, and not all eat at the same place. But what they’re doing is generating noise, and it’s extremely difficult to deduce a true pattern when you have a lot of noise. If the twitter users are all local to each other, the situation becomes much worse. Now, let’s look at a darker scenario. How many people would report Wendy’s or Taco Bell today, because of the recent reporting on these companies deciding to cut their fulltime staff because they don’t want to provide healthcare? Your system practically begs to be used for nefarious purposes. And the end result is increased time an inefficiency in medical systems already overtaxed.The entire point of doctors being involved at the beginning is that they can run tests that can rule out flu, or some other innocent cause from a more serious foodborne illness outbreak. They can’t easily be gamed, either. Any system has to start with them. You can’t crowd source food safety. 

  • Possibly a good idea