Blog 5: Data Collection, Analysis, and Node Centrality

In healthcare, web-based data entry has taken center stage regarding patient medical information.  Where in the past physicians and nurses entered information in paper charts, now there is a shift towards electronic health records (EHR).  By entering medical information during the encounter into an EHR, it mitigates recall bias and immediately makes the data available across treatment teams.

Image result for ehr collaboration

Moreover, because EHR data is entered in a standard format, it allows hospital coders greater accuracy and precision when reporting performance and quality statistics at the state and federal levels.

Image result for ehr standardization

Parallel to the consolidation of processes reported by Benítez et al., healthcare has also streamlined parts of data analysis and visualization.  Federal entities such as the Agency for Healthcare Quality and Research (AHRQ) have developed algorithms that allow organizations to analyze information reported to the Centers for Medicare and Medicaid Services (CMS).  This is significant because all hospitals that treat Medicare or Medicaid patients must report this data to CMS.

Image result for ahrq

In healthcare, quality measures used to assess quality and performance are constantly under question and the criteria used to gather and assess social network data should face similar scrutiny (Glance, Osler, Mukamel, & Dick, 2008).

It is through the process of verifying each question and measure that we arrive at a standardized instrument that can be used by others.   Standardization allows comparisons to be made across studies over time developing the body of knowledge needed to arrive at insights and correlations (Cicchetti, 1994).

Image result for standardization process

With the rise of social network analysis, do you believe the standardization process will become more and more salient?

The need for standardization points towards the inherent need in social network analysis for high quality data to understand centrality.  Once we have the data, we must understand the importance of node centrality measures as they are able to realistically represent human interactions.  In one study, social networks among students of a karate class, dolphins, and the neural network of a nematode were analyzed.

Image result for zachary's karate clubImage result for node centrality closeness betweenness eigenvector

Image result for c elegans neural network

The networks were studied for which measure of centrality really conveyed influence.  Their results identified key nodes as ones with high eigenvector and eccentricity centralities (Batool & Niazi, 2014).  Other factors of centrality considered include degree and closeness but these were not found to be significant.  The findings are consistent to our practicum exercise on crime networks as nodes high in eigenvector values were put forward as being high value targets.Image result for eigenvector centrality

The use of eigenvector values to understand influence is an intuitive one.  Alter nodes high in influence, power or resources being connected to an ego node confers power by associations.

Image result for eccentricity centrality

Eccentricity takes into account the maximal distance to other nodes from a central node  This is different from closeness because we are measuring how far a node is from one another, not how close.

Given the difficulty in realistically representing social networks, node centrality must be considered in the context of other factors.  For example, it was found in cases of complicated grief, node centrality and a node’s expected influence (EI) were both strongly correlated with network influence.  Here the EI was calculated as a measure of feelings of emptiness and emotional pain (Robinaugh, Millner, & McNally, 2016).  In fact, the EI was found to be more significant than node centrality in terms of influence.  It signifies the need to understand the data versus simply measuring centrality.

Image result for expected influence complex grief

The significance of EI points towards the work necessary in social network analysis to comprehensively explain behaviors and trends.  Node centrality is a strong starting point for understanding interactions, but the social network must be analyzed as a whole to explain observations.

Image result for beginner's mind

How could you practice beginner’s mind in social network analysis?


Batool, K., & Niazi, M. A. (2014). Towards a methodology for validation of centrality measures in complex networks. PLoS ONE.

Benítez, J. A., Labra, J. E., Quiroga, E., Martín, V., García, I., Marqués-Sánchez, P., & Benavides, C. (2017). A Web-Based Tool for Automatic Data Collection, Curation, and Visualization of Complex Healthcare Survey Studies including Social Network Analysis. Computational and Mathematical Methods in Medicine.

Cicchetti, D. V. (1994). Guidelines, Criteria, and Rules of Thumb for Evaluating Normed and Standardized Assessment Instruments in Psychology. Psychological Assessment.

Glance, L. G., Osler, T. M., Mukamel, D. B., & Dick, A. W. (2008). Impact of the present-on-admission indicator on hospital quality measurement: Experience with the agency for healthcare research and quality (AHRQ) inpatient quality indicators. Medical Care.

Robinaugh, D. J., Millner, A. J., & McNally, R. J. (2016). Identifying highly influential nodes in the complicated grief network. Journal of Abnormal Psychology.

Blog 4: Social Capital

Social capital can be understood under the framework of a social network (Lin, 1999).   It is an “investment in social relations with expected returns.” At the same time, social capital must be measured using “embedded resources in social networks.”  Following this logic, for us to understand social capital, we must study how it is developed within a social network as a “network asset.”

Image result for social capital network asset

Two perspectives exist on how social capital should be measured.  The first involves a focus on the individual.  It asks what are the connections that people have and how do these connections provide an advantage, profit or return for the individual?  The second asks how connections can provide these beneficial effects at the group level in the form of “collective assets?”  In general connections within social networks provided access to greater resources and information allowing the receiver to be more fit than they would be otherwise.  We previously discussed how even weak ties can provide these advantages (Granovetter, 1973).

Image result for individual collect social capital network asset

However, what happens when people lack social capital and a means to obtain it?  To explore this situation, we need to consider an investigation into the lives of the “Black Urban Poor” (Smith, 2005).  It was found that for poor African-American, they lacked access to the social networks which over time could potentially increase their social capital.  Without even the possibility of the ability to obtain social capital via networking, it was extremely difficult for them to obtain the resources and information they need to find suitable employment.

Given the potential benefits of social capital, and the deleterious effects of doing without, what can be done to improve social capital within our society?  Efforts undertaken to improve social capital should be undertaken with caution.  In the past, similar efforts were attempted under different names and many have failed to produce positive results (Portes, 1998; Portes & Landolt, 1996).

Our readings point towards the potential benefit of social capital at the individual- and group-levels but does not provide a viable way to apply this to many of the problems that we see in society.  For example, in health care, there may be people with a great deal of social capital, but managers may be ineffective in leveraging this resource.  Under this context my research project will attempt to reconcile social capital, and creativity into the application of valuable, cost-effective programs within the hospital setting.

It is without question that many people on the board of trustees for hospitals have a great deal of social capital.  In many circumstances, they received these appointments after proving themselves over an extended and significant period.  However, this social capital is rarely used to push forward programs that would improve the delivery of health care.

Image result for social capital

To help illustrate this point, consider the events that transpired after Cedar-Sinai Medical Center attempted to implement an Electronic Health Record (EHR) (Nembhard, Alexander, Hoff, & Ramanujam, 2009).  Without conferring with the hospitals staff and receiving their buy-in on how an EHR would improve efficiency and quality, an EHR was implemented.  Not surprisingly, after 1 week of use, most physicians got together and signed a petition to have the EHR removed.   What would you have done differently?  What did the implementation team fail to do?

Image result for cedar sinai medical center

The initial step should have been to obtain buy-in from the board of trustees.  Clear evidence indicating the efficiencies garnered by EHR use could then cause these trustees with high social capital to influence staff further down the hierarchy.  Then as the consensus towards EHR use became evident to more staff, they should have been included into presentations that demonstrated the use and benefits of an EHR directly.  Thus, buy-in from trustees and staff physicians would have increased the social capital of the EHR over time providing more favorable conditions for success.  In addition to this, members of the board of trustees could have been connected through weaker ties to more distant physicians for their input.  This clustered group composition with the addition of weak ties allows for greater creativity to surface while maintaining the expertise of the core group.  Could this be applied to other industries?

Image result for social capital fail

Consistent with the idea of social capital,  my small research project will attempt to study how the American Medical Association (AMA), and American Hospital Association (AHA)  attempt to exert their social capital by supporting politicians.  Specifically, the people that receive funds and their backgrounds will be put into context.  More so than the donation itself, knowing that the AMA and AHA place their support could have significant consequences.  Who are these people or organizations receiving support and is there wide variation in the size of donations?  How are their social networks structured?



Granovetter, M. S. (1973). The Strength of Weak Ties. American Journal of Sociology.

Lin, N. (1999). Building a Network Theory of Social Capital. Connections.

Nembhard, I. M., Alexander, J. A., Hoff, T. J., & Ramanujam, R. (2009). Why Does the Quality of Health Care Continue to Lag? Insights from Management Research. Academy of Management Perspectives.

Portes, A. (1998). Social Capital: Its Origins and Applications in Modern Sociology. Annual Review of Sociology.

Portes, A., & Landolt, P. (1996). Unsolved mysteries: The Tocqueville files II: The downside of social capital. American Prospect.

Smith, S. S. (2005). “Don’t put my name on it”: Social Capital Activation and Job‐Finding Assistance among the Black Urban Poor. American Journal of Sociology.

Blog 3 – Node Centrality is a Small World Network

A quote from Bonnie Frickson, helps us to understand the effects of centrality, “I argue that interpersonal processes vary with the kind of larger structural unit within which individual ties are embedded.”  Not all central nodes are created equal and to provide context, different forms of centrality must be defined.

Image result for node centrality closeness between

Degree centrality: is related to the number of connections (degrees) that a central node has.   Degrees provide more sources of resources, and more connections to receive and disseminate information.   The more connections (friends) you have the more likely you are to pick up on helpful information. Information can garner power.  Many connections also provide you with more  options for acquiring a resource, increasing the power of the central node (buyer).

Eigenvector Centrality: deals with connections to high scoring, influential nodes.  The more highly regard the connected node, the more power is afforded to the central node.  This is consistent with power by association.

Betweenness Centrality: is a central node that provides the shortest path between nodes.  These nodes are powerful to the extent that needed information is conveyed between nodes.  This bridging of weak ties is consistent with the “strength” of weak ties.   (Granovetter, 1973)

Closeness Centrality: is measured by the closeness  of a central node to other nodes.  Friends that are very close friends are more likely to help one another out thus providing power to each individual friend.

How can this understanding of node centrality be applied?   They can shed light on the types of connections that form throughout nature and society.  The concept is embodied in small-world networks.  These networks maximize connectivity while minimizing the resources needed for connections.  The power that is garnered from different forms of node centrality take on the form of efficiency in small-world networks.

Image result for small-world network brain

Efficiency is increased as inefficient connections are pruned.  This process is termed cortical plasticity as explained by Nobel Prize Laureate Eric Kandel (Kandel et al, 2012).  In essence, the human brain is born with many neural connections, but only neural connections that are efficient and productive get activated and reinforced over time.  Cortical plasticity is greater during early childhood and explains the ability of children to pick up new languages and learn new musical instruments very quickly.  It also explains how adults can still learn new functions throughout their lives, because some degree of plasticity persists.

By analyzing these small-world networks people have already derived a great deal of efficiency in other contexts.  For example the small-world network organization of the Fibonacci sequence has been applied to break complex codes in cryptography (Baldonia, 2009) .   It could also be applied to the design computers that maximize performance using the technology we have available.   In what other areas can you see a small-world network being applied?

It is up to us as social network researchers to understand intimately the structural differences between nodes and connection in order to elucidate how small-world networks can be applied more broadly.  To help us get there we need “reliability and validity of network measurement,” as a foundation for more and more complex network analysis (Robins).

Eric R. Kandel, MD; James H. Schwartz, MD, PhD; Thomas M., Principles of Neural Science, 2012

Maria Baldonia, Ciro Ciliberto, Giulia Cattaneo, Elementary Number Theory, Cryptography and Codes, 2009

Blog 2: Social Networks

The readings this week really helped me to look at networks from an enriched, more detailed perspective.   Research into networks deals with more than just personal attributes but with how people interact in relationships (Keim, 2011).

Image result for social network sample

The subset of lettered nodes in the picture above represent relationships that contribute to the whole.  The collection of these relationships provide us with social networks, where the whole is greater than the sum of its parts.  For example, it is the network that gives us social structure to understand what the norms of society may be.  Chances are, most of us follow these norms without even realizing because they are so ingrained into who we are.


At a more basic level, relationships can be thought of in the context of information exchange (Haythornthwaite, 1996).  It is information between actors, or nodes that causes feelings of friendship,  belonging and love to develop.  This flow of information can determine new connections that form, and also can effect the network as a whole.

Image result for social network information flow

Over time, the consequences of this flow can significant.  Take for example the idea of homophily (McPherson et al., 2001).  People have a natural tendency to group with those like themselves.  This can be seen in the relationships we have with friends, and in marriages.   Homophily allows for the possibility of isolation or cluster formation because of selective interaction.  As a consequence, perspectives can be narrower than they should be.  Consider what has happened to our political system as noted below. Image result for homophily

Contrary to generations past, homophily is making it more and more difficult to cross party lines.  The effects of homophily as our society becomes more and more interconnected is an opportunity for further research.  How will it affect our political system?

Social identity theory helps us understand the consequences of homophily.  As it is explained in the video, an ingroup and outgroup is formed causing differences between the groups to be amplified from bias and discrimination.  In effect, this causes the ingroup and outgroup to compete against one another rather than working together.  (A sad set of circumstances for our political system.)

The two textbooks assigned for reading this week helped me to understand the semantics behind the component parts and representations of networks and how they can be applied to research various social networks.   With my focus on health care administration, I especially appreciated how Robins used the example of snowball sampling to determine network boundaries.  This method was applied to determine if a relevant, representative  sample of decision makers was reached in the hospital setting.

Image result for health care electronic health record

As more and more hospitals implement electronic health records (EHR) increasing inter-connectivity between patients and physicians, as well as among physicians, I am curious to find out how this will ultimately affect social networks in health care.  Will different departments started working more closely together? What are some of the long-term consequences.  As of yet there are mixed reviews regarding EHR usage.  This is also an area for future research.


Blog 1: Connected

Reading the book Connected by Christakis and Fowler was a very eye-opening experience.  Never before did I image how interconnected we were to those around us.  I had previously heard that we were all connected to one another with at most six degrees of separation, but unaware of the level or impact of these connection.  According to the third degree rule, we significantly affect one another through three degrees of separation, with the effect getting weaker the more distant the connection.  This means that a friend, a friend’s friend, and a friend’s friend’s friend will all affect one another in tangible ways.  This influence persists even if the more distant friends never interact with one another directly.  To highlight this point, as it pertains to the social network below, Liz can affect Allen and Allen can affect Lisa, even if Liz and Lisa never have direct contact, she can affect her through indirect connections.  Moreover, the connection is a two-way street.

Image result for social network nodes

While simplistic, this transfer of influence can persist on a larger scale depending on the distance of nodes from one another and transitivity (Christakis and Fowler 2010, Social Network Sensors for the Early Detection of Contagious Outbreaks).  Distant connections, ones with few mutual ties, can serve as bridges allowing influence to reach new networks of connections.  In the picture above this could be Allen’s connection to Liz, allowing him to affect the highly transitive (many mutual connections) network on the left.  Increased transitivity allows members of the network to influence one another more readily.  In the case of a pathogen, if Allen was a carrier of a virus and exposed Liz, Liz could infect Emma and Shane.  Once Emma and Shane are infected, because of their high transitivity, the rest of their network could be infected.  This allows us to understand how pockets of disease can form, similar to what happened in Rockdale, GA.

Detailed interview transcripts:


network visualization of the outbreak

Other examples in the book used to illustrate connectivity include social network effects on patterns of obesity, voting, and smoking.

The insights gained from our study of social networks can be operationalized to help detect and prevent disease. Christakis and Fowler mentioned how vaccines could be targeted to highly connected individuals, at a central  hub of a network, to increase herd immunity versus immunizing the entire population.  This concept is shared by others. (Eubank et al.  2004, Modelling Disease Outbreaks in Realistic Urban Social Networks) Nodes that serve as a central hub can be used to detect the incidence of disease in a population. Targeting such individuals may also help in programs trying to change the behavior of a populations: criminals that regularly go in and out of jail could be more effectively persuaded to change their ways if people at the central hub demonstrated an aversion to crime or modified their own behavior.

Our discussion thus far has focused more on strong network connections but weaker connections also have a tremendous impact.  Weaker connections allow for the passage of information between transitive networks.  For this reason, weaker connections often are the source of new employment opportunity because they provide new information.  It has also been found through the study of the creators of musicals that a group with highly transitivity, coupled with more distant connections (providing different perspectives),  allows for the emergence of greater creativity as a whole increasing the  probability of success for the newly created musical.