Predictionof Epidemic Using Social Networks
Predictionof Epidemic Using Social Networks
The correct definition of an epidemic is a widespread outbreak of ahighly contagious disease within a community, at a particular time.Infectious diseases pose a threat to public health and a globalburden to national economies. Therefore, it is important that peopleunderstand the pathogenesis of various diseases to mitigate theiradverse effects to populations effectively. Traditionally, it is theCDC (Centre for Disease Control and Prevention) which has the legalmandate of determining a looming epidemic or an existing one. The CDChas the capability of determining the nature of an epidemic and itspossible trends over time. However, no certainty exists over theability to predict an outbreak before even the CDC has knowledge ofit. The unanswered question is the reason why a sociologist andphysician, Nicholas Christakis, studied the aspect and came up with arevelation of how people can predict epidemics using their socialnetworks.
The Network theory has in the past few years received attention as itpromised a framework that explains the patterns of epidemics. Here,the definition of a social network does not mean Facebook, Twitter,and the likes, but all different forms of face to face groups. Thisgroups could include and not limited to friends, towns, co-workers,relatives, among others. Some diseases spread by a form of socialcontagion like in the case of HIV where it is more likely to spreadamong persons with sexual ties. The human race has had these socialgroups since the beginning of time and therefore, it is not a newphenomenon. The reason for excluding online networks is because usersdo not necessarily know each other and two online friends cannotnecessarily produce robust and reliable data. Social networksdetermine the speed and the extent to which an epidemic can spread.In the recent past, data related to contact patterns has beencollected to define different epidemic situations. Therefore, thepurpose of this study is to identify the extent to which socialnetworks can predict the occurrence of an epidemic within acommunity.
Sociology is a diverse and invisible field which requires theacknowledgment of all aspects of the client. These mean that socialworkers are uncertain of the interventions that best work with thesociety and if they will be successful in controlling or managing thechallenges present. Sociology seeks to improve the status of thecommunity by finding appropriate ways of addressing the issues. Theadoption of social networks has been vital especially in predictingepidemics. Although social networks were used in the past, there hasbeen development and growth as a result of technological advancement.Networks consist of various relationships build between individuals.There is a high likelihood to reach all humanity through connections.The networks can be used to improve the world since it changes howepidemics are predicted. Social networks mean all forms of face toface groups such as coworkers, towns, and friends among others.Institutions such as Centers for Disease Control (CDC) take a lot oftime in the discovering and controlling epidemics as they have tocollect data and analyze them to determine the prevalence of aparticular disease. Alternatively, social networks have the abilityto speed up the process and lead to the early detection of epidemics.
Social relationships have positive and adverse effects on people’slives. They determine the maintenance and achievement of betterhealth. According to Ashida & Heaney (2008), a social network hasbeen defined as the webs of social ties where people live. Thenetwork possesses certain characteristics such as number, density,geographic, and homogeneity of network members (Christakis &Fowler, 2010). Through the social networks, the members can easilyinteract with each other since they have personal contacts and closegeographical proximities. People coexisting in a social network havetheir norms and values (Watts, 2003). The traditions influence thesocial aspect by changing their attitudes towards each other (Ashidaet al., 2008). Social comparison occurs as a result of the perceptionof similarities and dissimilarities in the network. In the case ofepidemics, the social network can quickly ascertain the problemswithin its members. The social support present in the networkprovides emotional, instrumental, appraisal and informationalsupport. Social connectedness promotes openness among individuals inthe network (Ashida et al., 2008). There is a possibility ofdetecting the outbreak of a contagious disease if individuals are ata center of a network. For example, a social network such as groupswould be more concerned in case of an infectious outbreak. The senseof trust in the network will force the members to share theirexperiences. Alternatively, responsible bodies such as CDC andnongovernmental organizations can collect the data with ease becauseof the central nature of the individuals within the social network.
Early knowledge is vital in managing epidemics especially contagiousones (Watts, 2003). Since most of the epidemics are complex andresistant to treatment in their late stages, it is important to haveprior knowledge on its unfolding (Bearman, Moody & Stovel, 2004).The presence of social network could be effective in theimplementation of strategies such as vaccination. The networkprovides information that can be monitored in real time. For example,a university health service may empanel sample of students who arefriends through their social network and have agreed to be passivelymonitored by practitioners (Christakis et al., 2010). The monitoringcould occur through visits to the health care facility. In thisscenario, a spike of a condition could translate as a warning of animpending outbreak. Public health officials find it easy to monitor asample of central people in the society because of the fewer burdens.However, in small institutions, central individuals may be vaccinatedsuch that they act as sensors in the community.
Watts (2003), argues that networks matter a lot in human life.Problems are represented as networks in various fields. The networkcan either influence behavior in a positive or negative ways.Collective behavior is similar to social network because people havetheir own rules of interaction. Individualism usually limits people(Watts, 2003). However, the decision-making process is unattainablesince individuals do not know what they want or need in the firstplace. It may be attributed to inadequate information or inability tohandle them. With the challenges of dealing with individual behaviorin decision making, people prefer to look at what other people aredoing (Watts, 2003). Social decision making is a component of asocial network. Information can be well aggregated through collectivedecisions. The fear of social comparison matters so much becausepeople have to pay close attention to others in the network. Throughthe process, they can have knowledge of the problem facing thesociety. The structure of the signaling network can drive or quash anepidemic. Haines, Beggs & Hurlbert asserts that social networkand support are distinct features (2008). Social networks comprise ofinterpersonal surroundings consisting of individuals and their socialrelationships with others. The connections present on the socialstructure influence the care and support received by people (Haineset al., 2008). With the joining of more members in the networksystem, there will be increased opportunity for them to get to knoweach other well. The companionship and friendship among theparticipants are beneficial since they can share relevant informationespecially when there is a problem such as a disease outbreak orhurricanes (Ashida et al., 2008).
People who have most connections to others in the society can befound and used to predict the rising trends in the larger group.Social network structures are powerful means of analyzing data andcould help to revolutionize the approach towards heath care andeducation. The study of the spread of HIV/AIDS in the sub-SaharanAfrica shows the importance of networks in predicting epidemics(Helleringer & Kohler, 2007). The spread networks connect thecore members of a group. For example, in sexual networks, there aresex workers and bridge population such as clients of the workers(Doherty, Padian, Marlow & Aral, 2005). The adoption of thesocial networks has helped in establishing a connection between therelationships and new HIV cases. Social workers can use theinformation and data from the groups to ascertain the percentage ofan epidemic in the population. The socio-centric network helps inidentifying the appropriate means of addressing the issue(Helleringer et al., 2007). From the study, sex workers have formed asocial network where they interact and inform each other regardingtheir clients. They have the ability to control the spread with thehelp of social support, and connectedness that exist between them(Ashida et al., 2008). Moreover, the analyses from the article showedthat there were varying results from the social network. The regionswith sparse population had a high prevalence of the epidemic becausethere were few individuals in the network (Doherty et al., 2005).People from these regions had a high rate of prevalence since theywere not aware of the infections in other parts (Bearman et al.,2004). Their network was regulated by the underrepresentation ofindividuals who participated in sexual activities. Social networksplay a great role in the distribution of HIV in the society. There isa possibility of being infected when one is close to the dense partsof the network. However, the infections will drastically decrease dueto the knowledge exhibited by a decline in health status amongindividuals. In small networks, people will tend to have longerrelationships which will directly translate to widespread of HIV(Helleringer et al., 2007). Individuals will take the assumptionsthat since the social network is sparsely populated, there will below rate infections. The social networks are beneficial to socialworkers because they will have the knowledge and expertise on theappropriate channels to follow when addressing the problem in thesociety (Christakis et al., 2010). Most people would be misled thatthe densely populated regions of the network have a high prevalenceof HIV, which is not the case because the sparsely populated faces agreat challenge (Bearman et al., 2004).
Christakis & Fowler propose that social networks are vitalcomponents in predicting epidemics in populations (2009). Theirargument is based on the connectedness logic. Online social networkssuch as Facebook and Twitter have been useful towards practitionersand government institutions such as CDC in handling epidemics (Saenz,2010). The platforms have been designed such that with a simple clickof a message, the information will automatically be forwarded toothers in the circle (Zhou, Zhao & Lu, 2015). Although some ofthe information on the social networks cannot be verified, there is ahigh likelihood that it will be accepted especially in a situationwhere many people have repeatedly mentioned it (Saenz, 2010). Socialreinforcement is an aspect of the social network. It implies thatpeople can only change and adopt new things if they experiencerepeated confirmations from their peers or neighbors (Zhou et al.,2015). Reinforcement ensures that the population within a givensocial network has information regarding everything happening aroundthem. Epidemics can be predicted through data mining from the socialnetworks (Saenz, 2010). For example, when people update their statusthat they have flu, it eases the process of ascertaining the spreadof the flu and proper control methods (Christakis et al., 2009).
Idealized networks are computer-generated and seek to define thebackground of disease transmission (Keeling & Eames, 2005). Theyhave been designed based on the distribution of people in a givenspace and their connections. Some of the most popular networksinclude random, lattice, spatial, small world and scale-free (Keelinget al., 2005). Random network is characterized by the inadequateclustering and homogeneity of people. The position of individuals onthe network is irrelevant since the connections are random in nature.Infection can only be spread through a fixed number of interactions(Kuperman, 2013). Random network has adopted a simple branchingprocedure when ascertaining the dynamics of diseases among apopulation. Lattices have been designed based on diverse assumptions.The model has positioned people on regular points on the grid. Theadjacent individuals are connected because of the presences of twodimensions (Kuperman, 2013). These mean that the contacts arelocalized in a given area. Unlike random network, the lattices arehomogenous and highly clustered because of the nature of theconnections. The individual level is part of the giant component ofthe population. Lattice model has used the contact process andforest-fire model in examining the spread of diseases (Keeling etal., 2005). When Lattice model is compared with other models such asrandom-mixing, they show a decreasing initial growth of infection.This can be attributed to the highly localized nature of theirtransmission and clustering (Keeling et al., 2005).
Additionally, the model comprises of “power-law scaling andself-organized criticality.” The frequency of the division of bothepidemic durations and sizes obeys power-law. Researchers haveadopted the lattice model aspect of scaling in observing the behaviorof the various diseases such as measles, mumps and childhoodinfections. Small-world networks have taken the weaknesses of randomand lattices (Kuperman, 2013). It has offered a path where anindividual can move from the unstructured connections in the randomnetworks and stiff arrangement of lattices. Small-world networksconsist of various characteristics such as short path lengths betweenindividuals and high clustering (Keeling et al., 2005). Thecharacteristics presented by the network means that there is a highlikelihood of infections occurring locally. On the other hand, theshort path implies that the epidemic may spread rapidly. A spatialnetwork is the most flexible type of idealized networks. In thespatial network, people have been positioned in a given localitywhere the connection of two individuals depends on connection kernel(Keeling et al., 2005). One can generate various networks by changingthe distribution of the kernel. Scale-free networks are heterogeneousin nature that is, few individuals have more neighbors while manypeople have few neighbors. There is a high degree of the spread ofdiseases because of the varying dynamics between individuals andneighbors.
The purpose of the study is to determine the significance of socialnetwork concept in predicting epidemics. With the increase in theworld population, the management of epidemics continues to be achallenge for the health practitioners and other stakeholders. Theconcept ensures that as people strive to use the social platform,they can be analyzed. In the modern world, traditional forms ofanalyzing and predicting epidemics are unsuccessful because thediseases have also changed. Group vaccination is only possiblethrough the social networks. The method is effective as a determinantsince its application and contribution to the health field is stillbeing felt and appreciated. When predicting trends of an epidemic, itis preferable to watch our most popular friends. There is a link thatexists between the social network and epidemiology of an infectiousdisease (Keeling et al., 2005). People have set many contacts in thepopulation through which they can pass an infection. Our most popularfriends are part of a social network which makes it easier for themto directly transmit diseases.
Before proceeding on the methodology used in this research, it isimportant to note this is a qualitative research. Hence, the studyfocused on secondary resources and advanced to collect and analyzethe findings. In this regard, this part of the paper concentrates onthe methodology and the results of the researches are discussedbelow. While there are various secondary resources used, the pioneersof the investigation were Christakis and Fowler who have beeninvolved in social networks for a many of years. Therefore, theirfindings and direction will never be underestimated in thedetermination of the social network`s prediction of epidemics.
This research is a qualitative secondary research, and therefore dataanalysis that will not be carried out. Instead, we use the secondarymethod of data collection. Secondary research provides a lot ofinformation and therefore it is most suitable for the research inepidemic prevention using social networks.
This section describes the collection of data for analysis. Thestarting point of any study is the technical evaluation because itgives information necessary for the research gaps before primaryresearch is carried out to fill the gaps (Liampttong, 2009).
Secondary Research Methodology
The secondary research involved studying the information from booksand articles on the topics of use of social networks to predictepidemics. Besides, several online materials were used to determinethe necessary information. In essence, this part involved readingnumerous books and articles on the concerned topic so as to gatherenough information. Various methods were used in the secondaryresearch methodology.
Firstly, the research involved studying how flu epidemic istransmitted from one person to another. However, they also indicatedthat such information applies to different areas such advertisementand political campaigns. In one of the studies, there was a selectionof some students who then to nominate their friends. The chosenstudents and their friends were then monitored closely for a givenperiod.
Secondly, a different experiment also simulated some people withtheir buddies. Basically, in this experiment was evaluating thefriendship ties and how the epidemic would flow from one person tothe next. Additionally, Christakis selected some people to identifytheir relationships and utilized the social structure that existedamong them to identify individuals who are central in the network.The "sensors," or the persons at the center, have mostfriends within a particular system. He then determined the path thatinformation and germs followed along the network.
Additionally, data was also analyzed by taking sampling Tweets on fluand comparing them with the incidences of the flu reported to thehospital. This step was done to determine whether there was anyrelationship between the two types of data. In other places such asthe United Kingdom, the comparison was made based on specific wordsthat the people had searched over the internet before the outbreak ofH1N1. This data was then analyzed to check for the relationshipbetween the number of type of tweets and the epidemic.
The methods thatwere used in the research involved random sampling since thepopulation of the respondents was significant. Emails were then usedto reach to then because of their numbers. Furthermore, sendingemails saves time.
Since thegathering of information was not first hand, the research in thisprocess might have some errors.
Secondary Research Limitations
The major limitation of this study was that of the books wereinaccessible and difficult to find. The articles were crucial in theinvestigation process. However, the most relevant articles wereavailable.
Summary of theResults
Christakis (2010), states that when in a social network and thirtypercent of the population is selectively immunized, then it isequivalent to the immunization of 100% of the population. Christakis`argument is that the population will come into contact with eachother through their social networks which will be crucial in aidingthe spread of the immunization from one person to the other.Similarly, an epidemic would spread by following such a path. Fromthe research which they did at the Harvard College, they found thatby monitoring the social networks among the selected students of thecollege and their friends, they could get a prior warning 16 days tothe epidemic. Central individuals are the members of the socialnetwork that have the most other members connected to them ideally,they have the most friends. According to the research whichChristakis did on the 714 students and their social ties, the centralindividuals are first affected by the epidemics before the otherstudents are affected. People at the edges of the network are likelyto suffer from an epidemic in the later stages. Epidemics spreadacross social networks. The social networks might include classmates,family members or friend. Christakis then concludes that for hisexperiment to work, there must be some form of interpersonalinteractions between the individuals.
Christakis and Fowler explain that this process could be applied todetermine whether a city will be affected by an epidemic. Forinstance, if a population in New York and a few of their friends in ashared network are affected by the flu, then the city should preparefor an epidemic of the flu.
Keeling & Eames (2005) also conducted an experiment to determinehow epidemics can be predicted using social networks. They found outthat emergent system could be used to establish which can factors areimportant epidemiologically. The system properties include distancefrom the seed, concurrent relationships and the number of suchrelationships.
Brown (2015) states that the research done by Tsou indicated thatthere is a very relationship between the number of reported cases offlu and the number of tweets on the flu. Specifically, the rate stoodat around 90%. Besides, when a comparison was between the predicteddata was against that of the hospital, the accuracy was close to 100percent. He continues to add that the research with similar apps suchas the “Flu near You” app can be able to predict with anaccuracy of some extent how the flu will spread in the area whichthey selected. However, there is a limitation of using informationfrom tweets since they are too generic and is not usually analyzed byprofessionals. Despite this, the social networks can be very usefulin the prediction of impending epidemics.
Schmidt (2012), addresses on pieces of research that had been doneon different outbreaks. For instance, he argues that during the 2009H1N1 outbreak, there were several searches on Google related to theepidemic. Such searches included "flu", "infection","hospital” and "influence". Additionally, before theoutput of ILI, there were several Google searches such as "swineflu," "sore head" and "cough night." Thescientists found that streams of Twitter tracked the diseases closelywith close probability. Whenever, there was an increased tweet abouta particular event, it signaled that the disease was about to affectthe many people in the country. What`s more about the findings isthat the predictions were always made between ten to fourteen daysbefore the surveillance of CDC could have indicated that there was anepidemic. Therefore, the social networks through the streams oftweets can help in the prediction of outbreaks.
In conclusion, the government should not rely on CDC and other bodiessimilar to only alone for it to get information on looming epidemics.But it should also use the prediction of epidemics using networksbecause accurate results are out two weeks earlier. The projectionmakes controlling any epidemic easier. However, there is more thatcan be found out by conducting thorough research. For instance, themethodology should study full networks and not just those at thecenter. Secondly, the research should consider all the mostfrequently used social media including WhatsApp and Instagram.
Coviello et al. (2010), make assumptions that the real infectionspass from one person to the other within the encounter network.Hence, they will be able to predict such disease if it is wellanalyzed. There is a direct proportionality between the risk of anode infected and its distance from the seed. If physical encounterdrives the infection from one person to the other, then theidentification of individuals at risk provided by friendship is notreliable. In the comparison of two networks, topologicalcharacteristics amplify the unpredictability which is in turndetermined by the randomness of the infection. Additionally, whenmonitoring the real in a relatively regular and infrequent manner,then the predicted infection on the encounter network can becorrected. As a result, the level of accuracy which obtained is high.To overcome the problems in the prediction of risks in the friendshipnetwork, then monitoring of the encounter system infection should beregular. When there is an efficient use of the random immunizationbudget, then there is a high probability that the disease will diewhen it is still in early stages. Furthermore, the final size ofinfection of the encounter network concerning randomization systemreduces significantly.
Ashida, S., & Heaney, C. A.(2008). Differential associations of social support and socialconnectedness with structural features of social networks and thehealth status of older adults. Journalof Aging and Health, 20(7),872-893.
Bearman, P. S., Moody, J., &Stovel, K. (2004). Chains of affection: The structure of adolescentromantic and sexual networks1. Americanjournal of sociology, 110(1),44-91.
Brown, J. (2015). Using SocialMedia Data to Identify Outbreaks and Control Disease. RetrievedDecember 21, 2016, fromhttp://www.emergencymgmt.com/health/Social-Media-Data-Identify-Outbreaks.html
Christakis, N. A., & Fowler, J.H. (2010). Social network sensors for early detection of contagiousoutbreaks. PloSone, 5(9),e12948.
Christakis, N. A., & Fowler, J.H. (2009). Connected:The surprising power of our social networks and how they shape ourlives. Little, Brown.
Coviello, L., Franceschetti, M.,Garc´ ıa-Herranz, M., & Rahwan, I. (n.d.). Predicting andcontaining epidemic risk using friendship networks.
Doherty, I. A., Padian, N. S.,Marlow, C., & Aral, S. O. (2005). Determinants and consequencesof sexual networks as they affect the spread of sexually transmittedinfections. Journalof Infectious Diseases, 191(Supplement1), S42-S54.
Haines, V. A., Beggs, J. J., &Hurlbert, J. S. (2008). Contextualizing Health outcomes: do effectsof network structure differ for women and men?. SexRoles, 59(3-4),164-175.
Helleringer, S., & Kohler, H.P. (2007). Sexual network structure and the spread of HIV in Africa:evidence from Likoma Island, Malawi. Aids, 21(17),2323-2332.
Keeling, M. J., & Eames, K. T.(2005). Networks and epidemic models. Journalof the Royal Society Interface, 2(4),295-307.
Kuperman, M. N. (2013). Invitedreview: Epidemics on social networks. Papersin Physics, 5,050003.
Saenz, A. (2010). How Social Networks Can Predict Epidemics AndControl the World (video). Singularity Hub. Retrieved 21December 2016, fromhttps://singularityhub.com/2010/09/21/how-social-networks-can-predict-epidemics-and-control-the-world-video/
Schmidt, C. W. (2012). TrendingNow: Using Social Media to Predict and Track Disease Outbreaks.Retrieved December 21, 2016, fromhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC3261963/
Watts, D. 2003. Epidemics andFailures. Chapter 6 (pp. 162-94) in SixDegrees: The Science of aConnectedAge. New York: W.W.Norton.
Zhou, C., Zhao, Q., & Lu, W.(2015). Impact of Repeated Exposures on Information Spreading inSocial Networks. PloSone, 10(10),e0140556.