ALTIMETRIK

ALTIMETRIK
Altimetrik is a rapidly growing organization with a vision to enable business innovation and heighten user experience by leveraging next generation technologies and design led engineering principles. This is achieved by incorporating user behavioral patterns into its solutions using accelerated design and deployment methods

Understanding Health Behavior using Social Network Analysis

Namrata Mishra, Research Analyst, Altimetrik

Altimetrik

Abstract
The health of a person has a vital impact on the collective health of his community and hence on the well-being of the society as a whole. In todays fast paced technology driven world, health issues are increasingly being associated with human behavior their lifestyle. Social networks have a tremendous impact on the health behavior of individuals these days. Many researchers have used social network analysis to understand human behavior that implicates their social and economic environments. This paper provides an algorithmic approach to analyze human behavior using social networks, which may have health implications. The results of this approach can be used by governing authorities for rolling out health plans, benefits and to take preventive measures, pharmaceutical companies can target specific markets, health insurance companies can better model their insurance plans.

Keywords
health behavior, social network analysis, crawler, directed graph, breadth first search, directed subgraph

Altimetrik

Introduction
In todays world, changing life style is affecting health behavior in our society, which impacts the health of individuals, community and society. Social Networks are immensely impacting the behavior of the society, which could be social, economic or health behavior. Hence, analyzing social networks becomes very useful. The MD Anderson Cancer Centre in Texas, USA, for example, moderated social networks such as The Cancer Survivors Network, Ask the Expert and The Anderson Network to better understand its patients experience and to improve patient care. This resulted in an increased understanding of how patients view their care experiences which helped MD Anderson to prioritize its scheduling processes.

Defining Health Behavior and its types

According to David Gochman, health behaviors are those personal attributes such as beliefs, expectations, motives, values, perceptions, and other cognitive elements; personality characteristics, including affective and emotional states and traits; and overt behavior patterns, actions, and habits that relate to health maintenance, to health restoration, and to health improvement. 1 So, we can say that any human action which has a direct or indirect impact on the health of a human being is health behavior.

Health Behavior can be categorized into four major types. First being Healthy behavior, which involves activities of individuals that keep them healthy, such as having a balanced diet, doing regular exercises etc. Second being Preventive health behavior, which involves activities undertaken by individuals for the purpose of preventing or detecting illness. It involves quitting smoking etc. Third is Risk behavior, which involves activities of individuals that cause sickness in them, such as consumption of alcohol, tobacco and having unsafe sex etc. Finally, there is Sick-role behavior, which involves activities of individuals that help them in recovering from a disease. It involves medical treatment etc.

Altimetrik

Understanding Social Network Analysis

Background Social Network Analysis (SNA) originates from the field of Social Science, Network Analysis and Graph Theory. Network Analysis is the mathematical analysis of problems which have a network structure and which are usually captured as a graph. A very early example of Network Analysis is from the city of Konigsberg (now Kaliningrad), where the famous mathematician Leonard Euler used a graph to prove that theres no path that crosses each of the citys bridges only once. Graph theory provides various methods and procedures to study and analyze graphs. These methods along with various other analytical and visualization tools, especially social networks, form a base for Social Network Analysis Social Network Analysis is not just about methods and techniques; it also provides a perspective of how a society functions. Instead of studying individual health behavior, Social Network Analysis helps to study the collective health behavior of a society Concepts of Social Networks Social Networks represent the social relationships among individuals, families, communities, organizations, regions, locations, villages and so on. Each of these units (which can play a dual or more roles in the social network) is represented as a node in the social network. Kinship, Friendship, Organizational relations, sexual relations, etc. are examples of some of the relations that exist in a social network. These relations are ties among the units. Social Networks Analysis Global Mapping of everything and their relations is referred to as Social Networks. In Graph theory, a graph is a set of objects which are connected with each other by some links. In mathematical terms, the connected objects are known as Vertices and the links that connect them are known as Edges. Social Networks can also be represented as graphs and they are known as Social Graphs.

Altimetrik

Social network analysis is a set of techniques, theories, concepts and methods that help in understanding social networks and relationships existing in these networks Social Network Analysis can be best represented as Relationship Technology. This technology is based on the description of the structure of network, relationships among units or actors and place of actors in social networks using network graphs, network matrices and mathematical analysis. Social network analysis focuses on the analysis of the relationship among the actors of the network instead of only analyzing attributes of the actors. The understanding of social network analysis can be used for various objectives. For instance, during the 2009 flu season, it was observed that H1N1 Youtube videos were viewed 2.6 million times, podcasts were downloaded 1.5 million times and the CDC Facebook page had over 50,000 followers. Social Media usage spikes coincided with outbreaks representing the publics use of information during the epidemic. Observing such patterns, enabled Government Agencies such as the Centers for Disease Control and Prevention (CDC) to help public share accurate information about health such as, H1N1 flu preparations and for engaging public during product recalls, etc.

Usage of Social Network Analysis

Social Network Analysis has been used in many domains other than social science. It has gained significant influence in the field of anthropology, biology, communication studies, geography, life sciences, history, information and communication, political science, computer science and development studies. Computer Scientists have used social network analysis to read web pages, internet traffic etc. In life science domain social network analysis can be used to study food chains in different ecosystems.

Altimetrik

There are various practical applications of social network analysis such as, Businesses use Social Network Analysis to improve communication in the organization or with partners and clients. Law enforcement organizations use social network analysis to identify criminal or terrorists networks with the help of communication information collected by them. Studies have proved the influence of social networks in the health care industry also. In a study, Thomas W. Valente, Ph.D., Professor of preventive medicines at the Keck School of Medicine of USC has mentioned the influence of social networks on health behavior.2 His study pinpoints the science behind social influence It talks about how important it is to understand the dynamics of a group or social network before designing interventions to bring about a change in their health behavior. Considering this influence of social networks on health behavior, I am proposing an approach for understanding health behavior using social network analysis.

Approach for understanding health behavior using social network analysis

Social networks can be represented as graphs which are known as Social graphs. In computer science terminology, a graph is a data structure which contains a finite set of ordered pairs (edges) that connect certain entities called nodes (Source: Wikipedia) These graphs can be directed graphs or undirected graphs. For example, in Facebook if X is friend of Y, Y must be a friend of X, So Facebook is an undirected graph. While in Twitter If X flows Y, Y need not to follow X, hence Twitter is a directed graph. This approach is proposing an algorithm considering Social Networks as Graphs.

Altimetrik

Say S (G) is a social graph of multiple locations.

Graph diagram [Figure] Figure1 S (G) In this approach, we are analyzing the health behavior of people in one location. So, from S (G) we will get a sub graph of a single location. As we know that to perform graph search there are two algorithms available, Breadth First Search and Depth First Search, we are using Breadth First Search Algorithm to get a sub graph of a single location.

Breadth First Search is a strategy for searching in a graph, where we first gain access to a node and inspect it. Then, we gain access to its neighbors and inspect them. This process continues till our search ends. (Source: Wikipedia) Depth First Search is a strategy for searching in a graph, where we begin with a particular node and explore as far as possible along each branch (Source: Wikipedia)

Altimetrik

Traversing S (G) using Breadth First Search algorithm gets a directed sub graph L (G) which is a location based graph. L (G) is a tightly coupled graph of location X.

Graph With Sub graph [Figure] Figure 2

Now that we have a location graph L (G), to narrow down our traversal we will get a sub graph of health communities, health blogs and users in the location. As we are analyzing health behavior of the people in that location, we need to know who the active users in the health communities and health blogs are, for which we are getting sub-graphs. Traversing L (G) using Breadth First search algorithm gets three directed sub graphs HC (G), HB (G) and U (G).

HC (G) is a tightly coupled sub graph of health communities or forums in the location X. HB (G) is a tightly coupled sub graph of health blogs in location X. U (G) is a tightly coupled sub graph of users in location X.

Altimetrik

Graph With one Sub graph and three more sub graphs of that sub graph [Figure] Figure 3 Traversing graph HC(G) using web based crawler identifies Number of users visiting health communities Number of users visiting health communities more than 10 times in a month Number of users visiting health communities less than 10 times in a month (U1) (U2) (U3)

Traversing graph HB(G) using web based crawler identifies Number of users contributing to health blogs Number of users contributing to heath blogs more than 10 times in a month Number of users contributing to health blogs less than 10 times in a month (U4) (U5) (U6)

Traversing graph U(G) using web based crawler identifies Number of users who drink Number of users who smoke Number of users who like health blogs/communities/pages Number of users who have health blogs as favorites Number of users who have taken health related surveys Number of users having at least 10 doctor friends Number of users who contributed in quit smoking blogs/surveys (U7) (U8) (U9) (U10) (U11) (U12) (U13)

Altimetrik

Processing information from U1 to U13 gives the following results Based on the results we determine that Locations with Healthy Behavior are those which have a higher percentage of Users contribution in health communities and health blogs, Active users in health communities and blogs and Users taking health surveys Locations with Risk Behavior are those which have a higher percentage of Inactive users in health communities and blogs and Users who drink and smoke Locations with Preventive Health Behavior are those which have a higher percentage of Users contribution in health communities and blogs, Users have at least 10 doctors contacts and Users contribution in quit smoking blogs and surveys. Demonstration using an example Let L (G) be a directed sub graph of a social network digraph S (G) with N nodes of location X. Using breadth first search algorithm on L (G) We derived directed sub graph HC (G) of health communities or forums in location X with m nodes. Directed sub graph HB (G) of health blogs in location X with n nodes. Directed sub graph U (G) of users in location X with 1000 nodes.

Altimetrik

Traversing graph HC(G) using web based crawler identifies 200 users visiting health communities 150 users visiting health communities more than 10 times in month 50 users visiting health communities less than 10 times in month Traversing graph HB(G) using web based crawler identifies 600 users contributing to health blogs 200 users contributing to heath blogs more than 10 times in month 400 users contributing to health blogs less than 10 times in month Traversing graph U (G) using web based crawler identifies 600 users drink 200 users smoke 450 users like health blogs/communities/pages 200 users have health blogs as favorites 100 users have taken health related surveys 400 users having at least 10 doctor friends 200 users contributed in quit smoking blogs/surveys Hence,

a a

a a

Percentage of users contributing in health communities and health blogs= (200+600+450)/3000 X 100 = 41. 66 Percentage of active users in health communities and blogs = (150+200+200)/ (200+600+1000) X100=30.55 Percentage of inactive users in health communities and blogs = (50+400)/ (200+600) X 100=56.25 Percentage of users who who have taken health surveys= 100/1000 X 100=10 Percentage of users who drink & smoke= (600+200)/ (2000 ) X 100=40 Percentage of users having contact=400/1000 X 100=40 at least 10 doctors

Percentage of users contributing in quit smoking blogs or surveys=200/(1000 ) X 100 = 20

Altimetrik

Based on the results we determine that

Average percentage of Healthy Behaviors in location X = (Percentage of users contributing in health communities and health blogs + Percentage of active users in health communities and blogs + Percentage of users taking health surveys)/3 = (41.66+30.55+10)/ 3 = 27.4 Average percentage of Risk Behaviors in location X = (Percentage of inactive users in health communities and blogs + Percentage of users who drink and smoke)/2 = (56.25 +40)/2 = 48.125

Average percentage of Preventive Health Behaviors in location X = (Percentage of users contribution in health communities and blogs + Percentage of users have at least 10 doctors contact + Percentage of users contribution in quit smoking blogs and surveys)/3 = (41.66 + 40 + 20)/3 = 33.88

A Case Study to explain the effectiveness of Social Network Analysis

The decision making process to introduce new vaccines into the National Immunization Programs is pretty complex and it involves numerous stakeholders who provide technical information, who mobilize finance, who implement programs and garner political support. When rolling out a new policy, it is important to understand how the policy makers interact and influence each other. A Social Network Analysis was done in Nigeria by interviewing federal and state government officials, officers of bilateral and multilateral partner organizations and other stakeholders such as health providers and the media.

Altimetrik

This analysis maps all the stakeholders involved in the policy making process, relationships between them and the pathway of influence (meaning who influences who and who is influenced by who) This facilitated an understanding that the policy actors who control finance allocation such as the Ministry `of Finance should be engaged in the policy making at an earlier stage itself and the state level actors who actually take care of the implementation could be engaged at the federal level to ensure the policy decisions taken centrally have a strong support for implementation locally. It also helped in the mapping of linkages and pathways to stakeholders who can influence critical decisions in the policy process. Visualization of analysis of health behaviors in different locations having different social networks with the proposed approach will be

Drawbacks of the proposed approach

Traversing graphs using breadth first search algorithm could be time consuming, which can increase the time complexity of this approach. Also, the approach is based on the data provided by users on social networks. This data may not always be correctly updated by the users and sometimes this data may not even be accessible owing to privacy, either of which could lead to inaccurate results.

Altimetrik

Business Usage of this approach

The Government can roll out health plans and benefits based on the health behavior of residents in their respective locations. They can also take preventive measures for any upcoming illness which may be because of the health behavior of residents in respective locations. The pharmaceutical companies can target specific markets based on the behavior of residents in different locations. Research Institutes can focus their research on issues which certain health behaviors can cause. They can also focus on the reasons for the health behavior of people in a particular location and define preventive measures in advance. Insurance Companies can come up with specific plans based on the health behavior of people. Design specific Policy plans for different types of profiles- healthy, unhealthy etc. The Results of this approach will be helpful for NGOs to identify the locations where Health Awareness Programs could be conducted. They will also be able to know the target audience for preventive health programs.

Conclusion
This paper illustrates an approach of using social network analysis to understand health behavior. The result of the proposed approach can be used by various governing and private authorities to grow their business and to contribute to improving the health of the society. Acknowledgements: This work is supported by Sriraman Raghunathan, Lead Architect, Altimetrik India Private Limited and Abhijeet Jaswal, Sr. Business Analyst, Altimetrik India Private Limited.

Altimetrik

Future work
In this approach, the parameters which determine the health behavior are very generic, such as users contributions to health communities and blogs or users who drink and smoke etc. In future, we can expand this approach to analyse more parameters such as users posts in social networks and users feedbacks on health surveys etc. which will lead to more accurate results. Also, in this approach, analysis of social networks is being done for three categories of health behavior only, which can be expanded to more categories. Moreover, this is a conceptual approach which needs to be implemented in future.

About the Author

Namrata Mishra is a Research Analyst at Altimetrik. Her areas of interest include Analytics, Social Network Analysis and their implications in the Healthcare industry. marketing-communication-in@altimetrik.com

Altimetrik

References
1. Karen Glanz, Jay Maddock: Behavior, Health Related Health Article, The Gale Group Inc., Macmillan Reference USA, New york, 2002. 2. Thomas W. Valente, Ph.D.: Social networks can be used to influence health behavior: study : Medical Press, Science 2012. 3. David Liben-Novell: An Algorithmic Approach to Social Networks. In: Department of Electrical Engineering and Computer Science in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Computer Science at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY ,June 2005. 4. Makism Tsvetovat, Alexandar Kouznetsov: Social Network Analysis for Start-ups, Publish by OReilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472. 5. Tianyi Wang, Yang Chen, Zengbin Zhang , Tianyin Xu, Long Jin, Pan Hui, Beixing Deng, Xing Li : Understanding Graph Sampling Algorithms for Social Network Analysis. 6. Silvio Lattanzi :Algorithms and models for social networks : Sapienza, Universit di Roma:Dottorato di Ricerca in Computer Science :XXIII Ciclo 2010 7. Thomas W. Valente, Ph.D., Peggy Gallaher, Ph.D.,Michele Mouttapa, Ph.D : Using Social Networks to Understand and Prevent Substance Use: A Transdisciplinary Perspective: SUBSTANCE USE & MISUSE Vol. 39, Nos. 1012, pp. 16851712, 2004. 8. Kirsten P. Smith, Nicholas A. Christakis: Social Networks and Health: Annu. Rev. Sociol. 2008. 34:40529. 9. Li, Charlene, and Bernoff, Josh. Groundswell: Winning in a World Transformed by Social Technologies, 2008. 10. H1N1 Web and Social Media Metrics Cumulative Data Report April 22, 2009 December 31, 2009, Division of eHealth Marketing (DeHM) National Center for Health Marketing (NCHM) Centers for Disease Control and Prevention (CDC). 11.http://www.jhsph.edu/research/centers-and institutes/ivac/resources/Health-Policy-Plan-Wonodi-socialnetwork-analysis-vaccine-decisionmaking-Nigeria.pdf

Altimetrik