Transcript Social Network Graphs & Data - eScholarShare

Using Social Network Analysis as a Tool to Evaluate
Medication Management in Ambulatory Care
Clare Tolliver, Dr. Andrea Kjos (mentor)
Drake University College of Pharmacy and Health Sciences
Introduction:
Results in Progress:
Social Network Graphs & Data:

Currently, three patients have returned consent forms to participate in the study.
 Social networks are created from person-toperson interactions and represent the
relationship between these persons.
 Graphically represents patterns within a social network

Data is being collected from their electronic medical records.
 Each dot represents one person in the network (node) and lines represent
connections between them.

Consent forms have been sent to twenty-five patients, and they will be added to d
collection as their forms are signed and received.
 These networks form in health care settings
between physicians, nurses, patients, and
other staff.
 Connections can be measured, based on the variable of interest:

Once data is collected, it will be used to create social network graphs to analyze t
communication patterns.
 Studying the communication patterns of these
clinical networks can identify the root of patient
safety issues, in this case medication
management.
Objective:

The goal of this project is to describe existing systems of medication management in
ambulatory care using network analyses methods.

Further, identify how network patterns may produce clues to how they link to patient
outcome characteristics and patient safety.

Results from the proposed study will determine the utility of a quantitative social network
analysis of organizational or ‘structural’ dynamics for evaluation of medication
management in ambulatory care.

Pilot Study: first time social network analysis methods will be used to study this type of
setting.
 Density - the number of people interacting
for each medication management task, will
provide information about network structure
and cohesion
Lessons Learned from a Pilot Study:
 Centrality - measured in terms of indegree and out-degree centrality scores,
who within the network are doing the most
communication initiation or reception, will be
used to determine relative prominence of
each person within the network

Support from clinic collaborators and staff is important.

The dynamics of a real-world practice setting, such as a clinic environment, may
changes in aspects during the course of study planning and data collection.

The researcher extracting data is purposefully an objective observer to remove p
for bias.

 Path length - a network property used to
calculate the distance between two persons
or “nodes” in a network. Average path
length can also be described as
‘betweenness centrality.’

Pilot Project Setting: Ambulatory Care Clinic in eastern Des Moines, IA

 Penn Avenue Internal Medicine


Roster survey instruments are a method used to collect data in social network analysis
when the research objective is to track who communicates within a given complete
network.
For this study, medication management will be defined as any task, communication, or
other exchange that links a minimum of two persons in the network regarding a patient’s
medication therapy.

Data will be collected by extracting information from electronic medical records of a
group of patients.

The network analysis will be described in terms of nodes (individual staff, providers and
patients) and ties (the number of communications between them).

Network analysis will focus on the interconnectedness (density) and the prominence
(centrality) of nodes in each network as is consistent with examining public health
systems.

Therefore, subjects become moving targets that are out of your control.

Subjects may have reasons to not want to enroll, move away or become lost t
follow up.
Difficulties arose with time spent contacting patients to participate; in the future, s
earlier with the process.

 22 staff members including physicians, nurses, pharmacist, and clerical.
Methods:
A longitudinal, roster survey will be used for collecting data in order to conduct network
analysis.
The nature of public health research is often conducted in an uncontrolled environ
Strategies for the Future:


The researcher abstracting data faces a “learning curve” to understand the wo
environment and idiosyncrasies of electronic medical records.
Ongoing Considerations:
Improving consent form intake process; if subjects can be contacted in person
time spent waiting for returned forms would be decreased.
To account for “learning curve” for navigating medical records, there is a need for
increased education for researchers abstracting data from the electronic system.
References:

Valente T. Social Networks and Health. Oxford: Oxford University Press, Inc.; 201

 However, it remains unknown how
vital other staff are in the
medication communication process.
Chang A, Schyve PM, Croteau RJ, O'Leary DS, Loeb JM. The JCAHO patient sa
event taxonomy: A standardized terminology and classification schema for near m
and adverse events. International Journal for Quality in Health Care. 2005;17(2):9

 How far are providers from each
other in terms of path length?
Knoke D, Young S. Social Network Analysis, 2nd Edition: Series for Quantitative
Application in the Social Sciences. 2nd ed. SAGE Publications, Inc; 2008.

Creswick N, Westbrook JI, Braithwaite J. Understanding communication networks
emergency department. BMC Health Services Research. 2009;9:247-255.
 Central nodes most likely will be
patient care providers.
 This may vary from patient to
patient and in terms of
seriousness of medication
related outcome/patient safety
communication.
Acknowledgements:
A special thanks to the project collaborators at Penn Avenue Internal Medicine and D
Ginelle Schmidt. This project was funded in part by the Drake Faculty Research
Awards Program.