Understanding a collaborative effort to reduce racial and ethnic disparities in health care: Contributions from social network analysis

Mathematica Policy Research, Washington, DC 20024, United States.
Social Science & Medicine (Impact Factor: 2.89). 07/2008; 67(6):1018-27. DOI: 10.1016/j.socscimed.2008.05.020
Source: PubMed
ABSTRACT
Quality improvement collaboratives have become a common strategy for improving health care. This paper uses social network analysis to study the relationships among organizations participating in a large scale public-private collaboration among major health plans to reduce racial and ethnic disparities in health care in the United States. Pre-existing ties, the collaborative process, participants' perceived contributions, and the overall organizational standing of participants were examined. Findings suggest that sponsors and support organizations, along with a few of the health plans, form the core of this network and act as the "glue" that holds the collaboration together. Most health plans (and one or two support organizations) are in the periphery. While health plans do not interact much with one another, their interactions with the core organizations provided a way of helping achieve health plans' disparities goals. The findings illustrate the role sponsors can play in encouraging organizations to voluntarily work together to achieve social ends while also highlighting the challenges.

Full-text

Available from: Marsha Gold
Understanding a collaborative effort to reduce racial and ethnic
disparities in health care: Contributions from social network analysis
q
Marsha Gold
a
,
*
, Patrick Doreian
b
, Erin Fries Taylor
a
a
Mathematica Policy Research, 600 Maryland Avenue, SW, Suite 550, Washington, DC 20024, United States
b
Department of Sociology, University of Pittsburgh, Pittsburgh, PA 15260, USA
article info
Article history:
Available online 23 June 2008
Keywords:
Managed care
Inequities
Health plans
Quality improvement
Network analysis
USA
Health care
abstract
Quality improvement collaboratives have become a common strategy for improving health
care. This paper uses social network analysis to study the relationships among organiza-
tions participating in a large scale public–private collaboration among major health plans
to reduce racial and ethnic disparities in health care in the United States. Pre-existing ties,
the collaborative process, participants’ perceived contributions, and the overall organiza-
tional standing of participants were examined. Findings suggest that sponsors and support
organizations, along with a few of the health plans, form the core of this network and act
as the ‘‘glue’’ that holds the collaboration together. Most health plans (and one or two
support organizations) are in the periphery. While health plans do not interact much
with one another, their interactions with the core organizations provided a way of helping
achieve health plans’ dispar ities goals. The findings illustrate the role sponsors can play in
encouraging organizations to voluntarily work together to achieve social ends while also
highlighting the challenges.
Ó 2008 Elsevier Ltd. All rights reserved.
Introduction
Many health care initiativesdincluding publicly spon-
sored efforts and public–private partnershipsdincorporate
a strong collaborative component to achieve their goals. In
recent years, quality improvement collaboration has
become a common strategy for improving health care.
One example involves the growing interest in rapid cycle
quality improvement efforts like those sponsored by the In-
stitute for Healthcare Improvement’s Breakthrough Series
(Berwick, 1989, 1998). Through collaboration, participants
learn about the variations in practice and identify potential
changes in care delivery processes that can promote change
(Kilo, 1998). Though most collaborations focus on general
quality improvement, some have gone beyond this to
address disparities in quality for diverse racial and ethnic
populations (Institute of Medicine, 2002; Landon, 2007).
While techniques for evaluating collaborative outcomes
are developing (Cretin, Shortell, & Keeler, 2004; Landon,
2007), less is known about evaluating collaborative
processes. Researchers generally rely on qualitative infor-
mation obtained through interviews with participants,
supplemented by numerical counts of events (e.g., meet-
ings, attempted changes). We complement this traditional
qualitative approach with formal analyses of a collaborative
structure and use social network analysis (SNA) to provide
additional insight on collaborative processes. Several exist-
ing studies have examined the organizational structure
(often through SNA) of coordinated service networks such
q
David Introcaso, then in charge of this evaluation for the Agency for
Healthcare Research and Quality (AHRQ), envisioned and supported net-
work analysis as a vital component of the evaluation. We are grateful for
the support of all participants in the National Health Care Collaborative
(NHPC) for their willingness to provide data essential to this work. This
study was funded as part of a contract between AHRQ and Mathematica
to evaluate the NHPC. At Mathematica, Judith Wooldridge provided valu-
able input throughout the evaluation. All views expressed in this paper
are those solely of the authors and not necessarily of any of the involved
organizations or NHPC participants.
*
Corresponding author. Tel.: þ1 202 484 4227; fax: þ1 202 863 1763.
E-mail addresses: mgold@mathematica-mpr.com (M. Gold), pitpat+@
pitt.edu (P. Doreian), etaylor@mathematica-mpr.com (E.F. Taylor).
Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
0277-9536/$ see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2008.05.020
Social Science & Medicine 67 (2008) 1018–1027
Page 1
as mental health networks, trauma centers, and services for
the aged (Bazzoli, Harmata, & Chan, 1998; Bolland &
Wilson, 1994; Goldman, Morrisey, Ridgely, Frank, Newman,
& et al., 1992). Directly examining the structure and func-
tioning of health care collaborative processes is the focus
of this paper.
The National Health Plan Collaborative
The National Health Care Collaborative (NHPC), formed
in late 2004, comprises large health plans from across the
United States and is sponsored by the Agency for Health-
care Research and Quality (AHRQ) of the U.S. Department
of Health and Human Services and by the Robert Wood
Johnson Foundation (RWJF). Sponsors hoped supporting
collaboration would encourage collective action and
engagement on reducing health care disparities and
encourage participating health plans to think creatively
about reducing disparities.
Together, the NHPC plans covered more than 76 million
people in the United States at its formation. FivedAetna,
CIGNA, Kaiser Permanente, United Health Group and Well-
Pointdprovide health care coverage nationwide in many
locations. FourdHarvard Pilgrim Healthcare of Massachu-
setts, HealthPartners of Minnesota, Highmark, Inc. in Penn-
sylvania, and Molina Healthcare, Inc. headquartered in
Californiadare regional firms. Participating plans became
involved in NHPC by attending meetings or being invited
to participate by NHPC sponsors or support organizations.
The NHPC decided to focus on diabetes because quality
measures are readily available and the condition affects
a large proportion of the population (and disproportion-
ately racial and ethnic minorities). Substantial disparities
are known to exist in diabetes care nationally (AHRQ,
2006), and the NHPC allowed health plans to explore and
confirm disparities within their own membership (NHPC,
2006) (For additional detail, see Gold et al., 2007.).
Nine participating health plans and six other organiza-
tions, as sponsors or supporters of the NHPC, were involved
in the first phase. The NHPC’s cosponsorsdAHRQ and
RWJFdeach actively participated in the effort’s activities.
Under contract to the sponsors, two primary ‘‘support
organizations’’dthe RAND Corporation (a large, nonparti-
san research organization) and the Center for Health Care
Strategies (a nonprofit organization focused on improving
quality and cost effectiveness of publicly financed health
care)dhelped coordinate efforts and provide technical
assistance. The Institute for Healthcare Improvement
(a nonprofit organization focused on improving quality of
health care) and a communications firms, GMMB, also
provided support.
The NHPC’s first phase was consistent with many health
care collaborations (Kilo, 1998). Over 2 years, the NHPC
held four in-person meetings of all participants. In addition,
support organizations held periodic conference calls with
individual health plans, helping each plan to measure
possible racial/ethnic disparities and to put interventions
in place to reduce disparities. Conference calls of all NHPC
participants took place periodically, and calls between the
sponsor and support organizations were held every few
months to discuss the future plans for the NHPC.
Fig. 1 shows a framework illustrating the pathways
through which the NHPC could influence work to reduce
racial/ethnic disparities. Participation was meant to help
enhance health plans’ commitment to reducing disparities.
These commitments translated into concrete actions that
could, over time, strengthen efforts to reduce disparities.
NHPC aimed to help plans develop and improve data on
disparities, and identify and implement pilot interventions
to reduce disparities. When ideas on data and interventions
developed, the NHPC encouraged sharing this information
between participants. The ultimate goal was communicat-
ing the collaborative results to organizations and others ex-
ternal to the NHPC. The NHPC is of interest because while
we know a great deal about the centralized control of
change (Goldman et al., 1992), we know much less about
how to leverage the voluntary energy of independent orga-
nizations to manage that change.
Methods
SNA was used to help understand collaborative
processes because it focuses attention explicitly on relation-
ships and ties
among the organizations. This is a critical sys-
temic feature beyond the attributes of network members
(Wasserman & Faust, 1994 ). SNA allowed us to examine
communication and collaboration across organizations.
The network data were collected between December
2005 and January 2006 (about 18 months after the NHPC
started) and were used to supplement ongoing, indepen-
dent and largely qualitative evaluation of the NHPC.
The specific network items were adopted from the much
replicated instrument developed by Van de Ven and Ferry
(1980). We selected items from the instrument that were
most relevant for assessing the NHPC and used modified
question wording specific to the NHPC’s context. Table 1
presents the network items, which ask each participating
organization to rate all other participants. Before asking the
network questions, each participating organization was
asked to provide feedback on the NHPC as a whole using
a set of nine items. These items give a benchmark that pro-
vides context for assessments of participating organizations.
Data collection was primarily by mail (with some orga-
nizations responding by fax). We asked the lead con-
tactdthe most senior person directly involved in the
NHPCdof each health plan, support organization, and
sponsor organization to provide information. To gain coop-
eration and help ensure truthful responses, we informed
participants at the outset that the identities of organiza-
tions would not be revealed in our results. The response
rate was 100%. Item-specific response rates also were gen-
erally high with the exception of one item, discussed below.
Analytical techniques
The SNA methods examined estimates of relations prior
to the formal collaborative participation, and relations
between organizations 18 months into the NHPC. The
structure of the whole network was delineated by using
generalized blockmodeling (Doreian, Batagelj, & Ferligoj,
2005) to provide the context for studying the participation
of each organization. We considered the extent to which
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–1027 1019
Page 2
Health Plan Interest in
Disparities
Join Collaborative,
Make Commitments
Become Part of Multi-plan
Effort
AHRQ and RWJF
Sponsorship
Complex Collective
Structure May Influence
Individual Firm Action
Enhanced Visibility,
Support
Identify and Pilot
Interventions Appropriate
to Their Context
Assess/Develop/Improve
Data on Disparities
RAND
Support
Exposure to
Other Plans
Participate in Collaborative
to Learn About/Develop
Effective Interventions
CHCS/IHI
Support
Exposure to
Other Plans
Plans Engage in Activities Related to
Reducing Disparities
Plans Share At Least Some Results,
Measures with Other Plans
Plans Continue to
Make and Measure
Improvements
in Disparities
Reduced Racial and Ethnic
Disparities as Quality
Improves
End of Formal Collaborative
Knowledge of How to
Reduce Disparities Grows
and Spreads
GMMB
RAND
CHCS
Fig. 1. Simplified conceptual framework of the NHPC to reduce disparities.
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–10271020
Page 3
relations were reciprocal, using measures of reciprocity
(Wasserman & Faust, 1994) and measures of organizational
standing (Doreian,1999) to see if a few organizations exces-
sively dominated the collaboration. Non-parametric
methods, using bootstrap (re-sampling) approaches
(Snijders & Borgatti, 1999), were used to provide robust
ANOVA tests that the network densities of the blocks iden-
tified by blockmodeling are significantly different. (Classi-
cal statistical inference is inappropriate given the
interdependence of the network observations.)
Limitations
The network we studied had a population of 15 organi-
zations. The small size is not a limitation in the sense that
we achieved full cooperation from all NHPC participants:
since this is the size of the whole system, descriptions of
the network structure are not compromised. (While each
organization’s lead contact for the NHPC completed the
network form, several organizations had multiple staff
work together on it.) However, our ability to generalize to
other larger systems may be restricted by the size of
NHPC. That responses were self-reported, despite the guar-
antee of confidentiality, also limits us because some organi-
zations may have been reluctant to provide negative
feedback on their peers. The analysis is restricted, by
design, solely to examining inter-organizational ties
regarding participation in the collaborative. These are
complex organizations and respondents may not necessar-
ily have had the full knowledge of all of their organization.
However, respondents were the senior leaders tasked by
the firms to represent firm views on the issues of concern
to the NHPC and were part of organizations that are
relatively formal in such designations. Thus, their
responses are the relevant ones for this study. Intra-
organizational examination of these large networks also
was not relevant. Finally, the data reflect only a single point
in time (and perceptions for a time period prior to the
NHPC’s formation) and therefore limit our insight into
changes in relationships over the course of the NHPC.
Study results
Overall perceptions of the NHPC
Participants were positive about the NHPC (Table 2). All
but one participating organization felt that the NHPC was at
least somewhat important to attaining organizational goals
(Question 1). In fact, 10 of the 15 organizations reported
that the NHPC was very important or crucial for achieving
organizational goals with regard to reducing health care
disparities. Overall, organizations felt that the NHPC has
carried out its responsibilities and commitments ‘‘to a con-
siderable extent’’ (Question 2).
Participants were positive about their own participation
in the NHPC (Question 3), and almost all respondents
reported the relationship between their organization and
the NHPC as productive and worthwhile (Questions 4 and
5). All organizations reported that they were satisfied
with their relationship to the NHPC to at least some extent
(Question 6).
Support and sponsor organizations reported they
changed or influenced NHPC activities more than plans
reported doing (Question 7). In contrast, health plans
were more likely than other organizations to say that the
NHPC had changed or influenced their organization’s activ-
ities, with six of the nine plans saying this happened to
a ‘‘considerable extent’’ or ‘‘great extent’’ (Question 8). All
organizations reported the payoffs of the NHPC were
reasonable relative to contribution (Question 9).
NHPC relationships in the context of collaborative goals
Strength of pre-existing ties
We queried respondents about relationships prior to the
NHPC between participants from different organizations.
While prior relationships are not a prerequisite for success-
ful collaboration, they do provide a baseline for assessing
information about the relations formed during the NHPC.
The findings reveal stronger pre-existing ties between spon-
sor/support organizations and participating health plans
than between the health plans themselves. Each of the
two sponsor organizations reported a prior acquaintance
with all or almost all other participants before the start of
the NHPC. The support organizations each had prior
acquaintance with at least 8 of the other 14 organizations,
though the key support organizations reported knowing
fewer organizations than other support organizations prior
to the start of the NHPC. Conversely, at least 10 organizations
Table 1
Network analysis domains and items
Pre-existing ties and their strength
In general, prior to the start of your organization’s participation in the
collaboration were you and were team personally acquainted with the
key people from the collaborative organizations listed? (answer yes or
no for each).
If yes: to what extent did your organization have an effective working
relationship with each of these organizations prior to the start of the
collaborative (to no or little extent; to some extent; or to considerable
extent)?
The collaborative process
During the past 6 m onths doutside of formal collaborative
meetingsdhow frequently have people from your organization who are
involved in the collaboration communicated or been in contact with
people in (each of) the organizations listed below (not at all, 1–2 times,
monthly, weekly, daily)?
To what extent has each organization carried out its responsibilities and
commitments involving disparities in regard to the collaborative during
the past 6 months (not at all important, somewhat important,
moderately important, very important, crucial)?
Which organizations provide good ideas for dealing with disparities at
meetings of the collaborative (no good ideas, some good ideas, many
good ideas)?
Perceived contributions of collaborative participants to action and change
Overall, how important was each organization’s work through the
collaborative in attaining the goals of your organization with respect to
disparities?
To what extent do you feel the relationship between your organization and
each of the organizations with respect to disparities is productive (to no
extent, to a little extent, to some extent, considerable extent)?
During the past 6 m onths, to what extent has each of these other
organizations changed or influenced the activities of your organization
with respect to disparities (to no extent, to a little extent, to some
extent, considerable extent)?
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–1027 1021
Page 4
reported a prior acquaintance with the sponsor
organizations and key support organizations before the start
of the NHPC.
We examined also reciprocal (mutual) awarenessdthat
is, organizations being personally acquainted with each
other (having reciprocal ties) prior to the NHPC as a founda-
tion for work in the NHPC. Organizations knowing each
other before entering the collaboration do not have to
spend as much time learning about each other as part of
the start-up costs of participation. Mutual awareness pro-
vides a limited platform for subsequent joint participation.
Fig. 2 displays these mutual awareness ties in the form of
a blockmodel, with three positions that include the core,
semi-periphery and periphery. The boundaries between
the three positions are marked by lines extending beyond
the outside square. The core is represented in the matrix
display by the first set of rows and columns, the semi-
periphery by the second set, and the periphery by the third
set.
A blockmodel was pre-specified (Doreian et al., 2005)as
follows (1) most of the potential ties will be present among
the organizations in the core (a complete block); (2) orga-
nizations in the core will be mutually aware of at least
one organization in the semi-periphery (regular blocks);
(3) organizations in the periphery will have few, if any,
mutual awareness ties, (null blocks) and (4) organizations
in the semi-periphery will not be mutually aware of each
other (null block).
The block structure in Fig. 2 shows that the pre-specified
block structure (with only four inconsistencies from the
specified core-periphery structure described above) fits
the data. Five of the sponsor and support organizations
plus one of the health plans (f8) are in the core. The
semi-periphery is made up of six health plans and one
support organization while the periphery has a single
health plan (f6). The only departures from the ideal block-
model are the absence of a mutual awareness tie between
s1 and s6 within the core and the presence of such a tie be-
tween f2 and f9 in the semi-periphery. The virtual absence
of ties between organizations outside the core may reflect
the fact that these health plans operate in a competitive
environment or in different regions. The densities of the
s1
s2
s3
s4
s6
f8
s5
f1
f2
f3
f4
f5
f7
f9
f6
s1
s2
s3
s4
s6
f8
s5
f1
f2
f3
f4
f5
f7
f9
f6
Black = Mutual awareness
White = No mutual awareness
Fig. 2. Reciprocated mutual awareness before the collaborative began. Note:
The rows represent how a given organization rates each of the organizations
listed in the columns. The diagonal of the figure is irrelevant since organiza-
tions were not asked to rate themselves.
Table 2
General perceptions of the National Health Plan Collaborative
All organizations (15 total)
1. Collaborative’s importance to organizational goals
Not at all important/a little important 1
Somewhat important 4
Very important 7
Crucial 3
2. Collaborative carried out its responsibilities and commitments
No extent/a little extent 0
Some extent 3
Considerable extent 9
A great extent 3
3. Organization carried out responsibilities and commitments to collaborative
No extent/a little extent 1
Some extent 1
Considerable extent 9
A great extent 4
4. Productive relationship between organization and collaborative
No extent/a little extent 0
Some extent 2
Considerable extent 9
A great extent 4
5. Time and effort spent with collaborative worthwhile
No extent/a little extent 0
Some extent 1
Considerable extent 7
A great extent 7
6. Overall collaborative satisfaction
No extent/a little extent 0
Some extent 1
Considerable extent 12
A great extent 2
7. Organization changed or influenced activities of the collaborative
No extent/a little extent 2
Some extent 4
Considerable extent 6
A great extent 3
8. Collaborative changed or influence activities of organization
No extent/a little extent 3
Some extent 5
Considerable extent 5
A great extent 2
9. Payoffs from collaborative reasonable relative to organization’s
contribution
a
No extent/a little extent 0
Some extent 1
Considerable extent 8
b
A great extent 5
a
One respondent did not answer this question.
b
Includes one respondent whose answer fell between ‘‘some’’ and
‘‘considerable.’’.
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–10271022
Page 5
parts of the network identified by the blockmodel (shown
in Fig. 2) are presented in Table 3.
Using the robust (against network autocorrelation)
ANOVA with a structural blockmodel option in UCINET
(Borgatti et al., 2002) yields a fit with R
2
¼ 0.428 where
the density in the core, 0.933, is significantly different
from 0 (p ¼ 0.001) and the density of the ties between the
core and the semi-periphery, 0.479, is significantly different
from zero (p ¼ 0.02) while 0.036 is not different to zero. The
number of observations is 210, the number of dyadic ties.
Also, with a separate test, the 0.933 is significantly different
from the 0.479 (p ¼ 0.001).
Within this broad structure, health plans varied
substantially in the number of organizations with which
they had a prior acquaintance, with the median plan
acquainted with only four other organizations before the
NHPC. Twelve of the 14 other organizations (including six
of the eight other plans) indicated a prior acquaintance
with one particular national plan. Participants most
commonly reported effective working relationships with
the sponsor and support organizations. Participants also
reported effective working relations with the key support
organizations on a relatively frequent basis. In sum, before
the NHPC began, health plans were more likely to have
effective working relations with sponsors or support
organizations than with one another.
The collaborative process
To understand the NHPC process, especially with
regard to participants’ communication and information
sharing, two network questions provided relevant infor-
mation: the extent of communication and providing useful
ideas.
Communication.
While formal meetings involving NHPC
organizations were a key activity, we also wanted to see
if participating organizations communicated with each
other outside of these formal gatherings and asked about
these informal communications. We fitted a blockmodel
to this relationship with only a core position and a periph-
ery position (because f6 became more linked to the rest
of the network). (This periphery is a re-labeling of the
previous semi-periphery.) The results are shown in Fig. 3
for the unique optimal partition. There is a small amount
of missing data for this relation, marked with diamonds
in Fig. 3, but treating them as null ties or as actual ties
does not change the number (10) of inconsistencies
between the idealized and fitted blockmodel. In contrast
to the blockmodel fitted to the pre-collaborative mutual
awareness ties (see Fig. 2), we now find that one health
plan (f6), which was previously in the periphery, interacts
with some of the other organizations, a consequence one
might presume of the participation of this plan in the
Table 3
Densities identified in the blockmodel of mutual awareness ties
Core Semi-periphery Periphery
Core 0.933 0.479 0.000
Semi-periphery 0.479 0.036 0.000
Periphery 0.000 0.000 0.000
s1
s2
s3
s4
s5
f2
f9
s6
f1
f3
f4
f5
f6
f7
f8
s1
s2
s3
s4
s5
f2
f9
s6
f1
f3
f4
f5
f6
f7
f8
Weekly communication
=
=
=
=
=
Monthly communication
Communication 1-2 times
No communication
Missing data
Fig. 3. Frequency of communication between collaborative participants. Note: The rows represent how a given organization rates each of the organizations listed
in the columns. The diagonal of the figure is irrelevant since organizations were not asked to rate themselves.
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–1027 1023
Page 6
collaboration. Again, the network densities differ systemat-
ically, consistent with the specified blockmodel. The
densities for the four blocks (shown in Fig. 3) are presented
in Table 4.
The robust ANOVA has an R
2
¼ 0.310 with n ¼ 210, the
number of dyadic ties in the network. Using the internal
periphery ties as the omitted category, the density 2.524
is shown to be significantly different from 0.179
(p ¼ 0.0002), the density 1.571 is significantly different
from 0.179 (p ¼ 0.001) and the density 1.286 is significantly
different from 0.179 (p ¼ 0.001). When the internal core ties
form the omitted category, a robust ANOVA reveals that
1.571 is significantly different from 2.524 (p ¼ 0.0024) and
1.286 is significantly different from 2.524 (p ¼ 0.0112).
The structure of communications outside of formal
collaboration meetings has a core-periphery form but the
composition of the organizations in the core and in the
periphery changes slightly. Four of the five sponsor and
support organizations of the core for the pre-collaborative
awareness ties remain in the core but one (s6) drops out.
Two of the health plans (f2 and f9) are now in the core of
the network while one (f8) is not there for this relation.
Fig. 3 shows s6 is not in this core, because it communicated
a little less with core members than did two of the health
plans (f2 and f9), and shows that support organizations
report communicating extensively with each other. Com-
munication in the core of the network is the most ‘‘dense’’
part of the communication network. Some communication
also occurred between the two key support organizations
and the two health plans in the core for this relation. This
result is consistent with what we understood to be the
way that the NHPC worked in the time period studieddan
approach involving extensive consultation between
individual support organizations and plans, and also
among the support organizations in order to coordinate
efforts. Outside of formal NHPC meetings, only a few plans
reported communicating with other plans, with regional
plans reporting a large share of such communications.
Three plans (two national and one regional) reported no
communication with any other plan outside of NHPC
meetings.
Providing good ideas.
Most organizations were rated by
others as providing the NHPC with at least ‘‘some good
ideas.’’ Plans and non-plan organizations alike often rated
the two key support organizations and the two sponsors
this way. In addition, one national plan and one regional
plan were identified by over half of the other organizations
in the NHPC as providing many good ideas. Conversely,
two national plans were identified by nearly half of the
other organizations as providing no good ideas. One plan
saw only three other organizations (all support organiza-
tions) as a source of good ideas. Six of the participating
organizations saw all other organizations as sources of
good ideas.
Perceived contributions of collaborative participants to
action and change
While understanding the NHPC process is important
(for its own participants to understand how the group
is functioning and for others interested in how collabora-
tions in general can work most effectively), the NHPC’s
ultimate aim is to bring about concrete changednamely,
organizational action towards reducing racial and ethnic
disparities among participating health plans. We sought
to assess whether any organizations were particularly
important or influential (or not) to the NHPC. Three net-
work items provide information on the perceived contri-
bution of other participating organizations to a given
organization’s actions and goals: each organization’s
importance, productivity, and influence relative to other
organizations.
Importance of others.
Representatives of each participating
organization were asked to rate the importance of others
in the NHPC for the ability of their own organization to
attain its own goals with respect to reducing disparities.
Other organizations can be seen as important even if
they do not necessarily directly interact because they
can bring knowledge to the NHPC that spreads via the
NHPC relationships. We fitted a center-periphery block-
model (data not shown). All of the sponsor and support
organizations are in the core (other than s5) and they
are joined by three of the health plans (f1, f2, and f4).
With regard to importance of other organizations for
each organization, the network’s core has become larger
revealing a stronger integration of these three plans
with the support organizations in the core. In addition,
both the block representing the ties from the core to the
periphery and the block for the ties from organizations
in the semi-periphery to those in the core are denser.
The density of ties between organizations in the semi-
periphery is considerably less, suggesting less integration
between these organizations.
Health plans generally rated non-plan organizations as at
least moderately important in helping the plans attain their
disparities-related goals. In addition, two of the health plans
(f1 and f2) viewed the remaining health plan organizations
as important. However, the other six health plans did not
identify other plans as important to their organizational
goals. In general, plans rated the support organizations as
important because sponsor and support organizations
were directly involved in helping plans with their work on
disparities.
Sponsor and support organizations generally rated each
other as at least moderately important to attaining organi-
zational goals and such perspectives in fact tended to be
reciprocated among these organizations. Sponsor and
support organizations tended to rate plans as at least
moderately important to organizational goals. Thus, all
three groups perceived that sponsors and support organi-
zations were at least moderately important to the success
of the NHPC while plans were less likely to perceive other
plans as important to them.
Table 4
Densities identified in the blockmodel of communication
Core Periphery
Core 2.524 1.571
Periphery 1.286 0.179
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–10271024
Page 7
Productive relationships with others.
When asked about the
productivity of relationships with other NHPC participants,
most organizations saw their relationships with others as
productive at some level (i.e., at least ‘‘to a little extent’’).
Collaborative participants most frequently reported con-
siderably productive relationships with the key support
organizations and the sponsor organizations and, to a
lesser extent, three of the other plans.
Fig. 4 provides a network graphic representation of
organizations reporting considerably productive relation-
ships with other organizations participating in the NHPC.
(This figure was created by using a spring embedder, with
slight adjustment.) Sponsor and support organizations are
shown as rectangles and plans as ovals. Lines between
organizations with a single arrow represent one organiza-
tion reporting a productive relationship with the other
organization (with the receiving organization being the
one with which the relationship is considerably produc-
tive). Lines without arrows indicate that both organizations
reported considerably productive relations with one
another. Most plans are viewed as offering considerable
productive relationships by only a few other organizations
in the NHPC.
Influence of others.
Collaborative participants were asked
to assess the extent to which other organizations in the
NHPC changed or influenced the activities of their own
organizations relative to disparities. Sponsor and support
organizations are reported to have the most influence on
other NHPC members (one of the key support organizations
and one of the sponsors have the most influence). Only
one organization (a regional plan) reported no external
influence from any other participants. With this one
exception, all NHPC organizations say they have been influ-
enced by other participants.
Overall findings of organizational standing
Given the NHPC is a collaborative venture, we wanted to
see if one or more organizations unduly dominated the
NHPC. Also, we wanted to examine whether there were
organizations contributing an unusually small amount to
the collective enterprise. We constructed a general index
of relative standing for each of the NHPC participants (Dor-
eian, 1999). The features of this measure of standing are (1)
organizations have greater standing if they receive more
‘‘nominations’’ identifying them as influential or impor-
tant; (2) greater standing is associated with receiving
f1
f5
f2
f9
f7
f4
f8
f3
s3
s5
s6
s1
s4
s2
f6
= Sponsor or support organization
= Plan
Fig. 4. Network diagram showing reports of considerably productive relationships between collaborative participants. Note: responses to this question included
to a considerable extent, to some extent, to a little extent, and to no extent. Ties shown in the figure reflect relationships that are productive to a considerable
extent. Ties with a single arrow mean that one organization rated the other organization (receiving the arrow) as considerably productive. Ties with no arrows
indicate that both organizations reported considerably productive relationships with one another.
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–1027 1025
Page 8
nominations of greater strength (e.g., influence is ‘‘consid-
erable’’ rather than ‘‘little’’), and (3) organizations have
greater standing if their nominations come from other
organizations with high standing. We analyzed relative
standing for several dimensions (importance, responsibili-
ties and commitments, productive relationships, source of
good ideas), but report only the overall findings.
The analysis of relative standing revealed that the two
key support organizations and two sponsors generally
have the highest standing in the NHPC. The same did not
hold for the measure of standing related to source of
good ideas; some plans rated higher than non-plan
organizations on this measure. In addition, one of the
national plans consistently had relatively high standing
across several dimensions, as did one of the regional plans.
Most notably, there are no outliers in the distributions of
relative standing, as defined by data points more than 1.5
times the interquartile range beyond the quartiles. In other
words, even though participating organizations vary in
their standing, no organizations stand out at either extreme
in contributing to the NHPC.
Discussion
Summary of findings
Organizations are necessarily interdependent and these
interdependencies generate complex networks with coop-
erative and competitive ties. The network analyses high-
light the central role of the NHPC’s sponsor organizations
and primary support organizations as part of the core of
the network for all of the studied relationships (with s5
as an exception). During the first few years of the NHPC,
sponsors and primary support organizations had the most
contact with participating plans and formed the primary
pathways linking participants. They were visible and active
participants in the NHPC process that acted as the ‘‘glue’’
holding the NHPC together during its first phase. The
primary sponsor and support organizations also engaged
in a substantial amount of contact with one another. The
prestige the sponsors brought to the NHPC also was a val-
ued contribution for plans.
Consistent with our observations of NHPC meetings and
conference calls, SNA also confirmed that plan-to-plan
interaction was not very prevalent within the NHPC’s first
phasedand much less prevalent than plan-to-non-plan
relations. But the SNAalso reveals that health plans were im-
portant to the NHPC, despite their limited interaction with
one another. Two to three plans were in the core in most
of our analyses. The NHPC processes led plans to have an in-
fluence on one another even if they did not directly interact.
Organizations’ ratings of whether the NHPC is produc-
tive and worthwhile and whether it yields a reasonable
payoff compared with the level of organizations’ contribu-
tions are all fairly favorable and do not appear to vary
greatly by organizational standing.
Interpretation
This study reveals that health plans rarely communi-
cated with each other 18 months into the NHPC’s first
phase. Our findings also indicate that an absence of
communication can be offset by contact and communica-
tion occurring through other pathways involving the sup-
port and sponsor organizations. Interactions between
organizations consume time: it would be inefficient if all
organizations in the collaborative communicated with
each other. Moreover, given the competitive nature of the
U.S. health care industry, unless plans operate in different
markets (as with regional plans) or have some business
imperative for additional collaboration, they may well limit
their direct contacts.
The findings of our research are consistent with the
different motivations of participating organizations in the
NHPC. Several plan staff reported in our qualitative inter-
views that, primarily, they wanted to learn what other
plans were doing in the area of racial and ethnic disparities.
Fewer sought specific support in making changes. Further,
although plans did not necessarily say that they were reluc-
tant to share information, their internal clearance
processes made it clear that release of plan-specific infor-
mation is an important threshold decision. These motiva-
tions and constraints help account for the NHPC’s
network structure and why sponsors and support organiza-
tions’ played a crucial role in the way the NHPC functioned
during its early phase. Moreover the presence of neutral,
external sponsors and the backing of a governmental
agency emphasized the importance of work on reducing
disparities for participating health plans.
The limited communication, however, points to a poten-
tial weakness in the network structure: if the collaborative
ends without greater communication among the non-core
organizations, the absence of the core will leave a very
sparse network. Thus, the results show that centralized
support can engage organizations in joint work but also
that there is a downside risk to sponsors that the ongoing
ability to collaborate might depend on ongoing outside
support. Those who invest in collaboratives need to think
about both short- and long-term goals and whether they
can be pursued if a collaborative ends. Activities that seek
ultimately to encourage change within organizations may
be easier to sustain than those that require ongoing interac-
tion across organizations.
Practical application
While our SNA and qualitative analysis address similar
questions, they do so in complementary ways. The NHPC
experience provides at least two examples illustrating the
practical application of this complementarity.
The first involves generating information useful in
enhancing the group process via feedback. Through our
qualitative analysis, we knew the sponsors/support organi-
zations provided the main focus for communication and
interaction in the NHPC and the interviews provided
insight into the competitive environment and reasons for
potential hesitancy plans had in communicating openly
with one another. SNA provided a concrete and systematic
image of the NHPC’s operation. Because results were shared
with plans, this aspect of the NHPC was more accessible to
them. One participating plan representative, for example,
noted at an NHPC meeting that the representative was
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–10271026
Page 9
asked by superiors after reviewing these SNA findings why
there was so little direct collaboration in a collaborative
exercise. Sponsors used the ensuing discussion to raise
possible ways to encourage more interaction as the NHPC
proceeded. If this does not happen, the chances for success-
ful outcomes will be reduced.
The second example concerns SNA providing systematic
ways of seeing whether perspectives are reciprocal across
participants in a collaborative process. In studying organi-
zations, there is the risk of a natural tendency for represen-
tatives to promote their organizations in positive terms.
Some participants also may have perspectives on others
that may not necessarily be consistent with reality. SNA
provides valuable tools for generating independent insights
that can be used to address such concerns.
Research contribution
This study originated out of concern that evolving tech-
niques for evaluating the growing number of learning
collaboratives on quality improvement could benefit from
more formal analyses of the processes they seek to influ-
ence (Cretin et al., 2004; Landon, 2007). Our study
employed the social network techniques previously applied
to organizational structures of other health-related
settingsdsuch as mental health networks, trauma centers,
and services for the aged (Bazzoli et al., 1998; Bolland &
Wilson, 1994; Goldman et al., 1992) d to understand better
the way the structure and process of a learning collabora-
tive worked. The results of our study demonstrate the
potential value of these techniques in completing a multi-
method and multi-focus evaluation. SNA, in particular,
enhances the ability to visualize and make concrete the
interactions within a network or organizational structure.
For the future, the challenge will be to examine more
closely how these processes and interactions change over
time and particularly how they influence outcomes.
References
Agency for Healthcare Research and Quality. (December 2006).
2006 National healthcare disparities report. AHRQ publication
no. 07-0012. Rockville, MD: U.S. Department of Health and Human
Services, AHRQ.
Bazzoli, G. J., Harmata, R., & Chan, C. (1998). Community-based trauma
systems in the United States: an examination of structural develop-
ment. Social Science and Medicine, 46(9), 1137–1149.
Berwick, D. (January 5, 1989). Sounding board: continuous improvement
as an ideal in health care. New England Journal of Medicine, 320(1),
53–56.
Berwick, D. M. (April 15, 1998). Developing and testing changes in care
delivery. Annals of Internal Medicine, 128(8), 651–656.
Bolland, J. M., & Wilson, J. V. (1994). Three faces of integrative
coordination: a model of interorganizational relations in
community-based health and human services. Health Services
Research, 29(3), 341–366.
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). UCinet for Windows:
Software for Social Network Analysis. Harvard, MA: Analytic
Technologies.
Cretin, S., Shortell, S. M., & Keeler, E. B. (2004). An evaluation of collabo-
rative interventions to improve chronic illness care: framework and
study design. Evaluation Review, 28(1), 28–51.
Doreian, P. (1999). Organizational standing as social capital. In R.
Leenders, S. M. Gabbay, & Dorrecht. (Eds.), corporate social capital.
The Netherlands: Kluwer.
Doreian, P., Batagelj, V., & Ferligoj, A. (2005). Generalized blockmodeling.
New York: Cambridge University Press.
Gold, M., Taylor, E. F., Gruene Segersten, K., Doreian, P., Coughlan, J., &
Lipson, D. (December 2007). Evaluation of a learning collaborative’s
process and effectiveness to reduce health care disparities among
minority populations. Final report prepared for the agency for health-
care research and quality. Washington, DC: Mathematica Policy
Research, Inc..
Goldman, H. H., Morrisey, J. P., Ridgely, M. S., Frank, R. G., Newman, S. J., &
Kennedy, C. (1992). Lessons from the program on chronic mental
illness. Health Affairs, 11(3), 51–68.
Institute of Medicine. (2002). In B. D. Smedley, A. Y. Stith, & Alan R. Nelson
(Eds.), Unequal treatment: confronting racial and ethnic disparities in
health care. Washington, DC: National Academies Press.
Kilo, C. M. (1998). A framework for collaborative improvement: lessons
from the institute for healthcare improvement’s breakthrough series.
Quality Management in Healthcare, 6(5), 1–13.
Landon, B. E. (March 1, 2007). Improving the management of chronic
disease at community health centers. New England Journal of
Medicine, 356(9), 921–934.
National Health Plan Collaborative. (November 2006). National Health
Plan Collaborative: phase one summary report. Princeton NJ:
Center for Health Care Strategies. Available from: http://www.
chcs.org/NationalHealthPlanCollaborative/index.html. Accessed 14.
12.07.
Snijders, T. A. B., & Borgatti, S. P. (1999). Non-parametric standard errors
and tests for network statistics. Connections, 22(2), 61–70.
Van de Ven, A. H., & Ferry, D. L. (1980). Measuring and assessing
organizations. New York: John Wiley and Sons.
Wasserman, S., & Faust, K. (1994). social network analysis: Methods and
applications. Cambridge, England: Cambridge University Press.
M. Gold et al. / Social Science & Medicine 67 (2008) 1018–1027 1027
Page 10
  • Source
    • "Finally, there was the suspicion that one organization largely unknown prior to the collaborative effort. This organization (or possibly more than one organization) occupies Position 3. Reflecting these arguments, the specification used in [15] is shown in Table 3. This pre-specified model was fitted with 3 clusters and the rows of the matrix were permuted so that members of the same position (cluster) are grouped together. "
    [Show abstract] [Hide abstract] ABSTRACT: Glossary Definition of the Subject Introduction Conventional Blockmodeling Generalized Blockmodeling Examples of Generalized Blockmodeling Future Directions Bibliography
    Full-text · Article · Nov 2013
  • Source
    • "One type of NA, organizational network analysis (ONA) examines the relationships between groups or organizations , including uncovering and examining patterns of behavior; the situations under which these behavior patterns arise; and the consequences of specific patterns on group behavior. ONA has been used to examine collaboration and information sharing between public health agencies and health plans626364 . Social network analysis (SNA) follows the same premise, but concentrates on understanding the relationship between the individuals (e.g., providers). "
    [Show abstract] [Hide abstract] ABSTRACT: Health information technology has been shown to influence the communication patterns of healthcare providers. The goal of this study was to learn more about how healthcare providers communicate and exchange patient clinical information during patient handoffs (transfers) between units in an acute care setting. Convenience sampling was used to select five patient handoffs. Questionnaires were distributed to providers identified through observation and snowball sampling. Social network analysis methodology was used to develop sociograms of the emergent communication patterns and identify the role of individual providers in the handoff process based on the number of contacts with other providers and incoming and outgoing communication activity. Individual handoff network size ranged from 11 to 20 providers. Participants were asked to describe the method of communication they used to access or share clinical information with other providers, their preferred method of communication; their satisfaction with the available options; and their suggestions for how the process could be improved. The network patterns that emerged uncovered the overlapping use of synchronous and asynchronous communication methods (verbally via phone or in person; or written via paper charts and/or an electronic records). No particular professional group dominated or coordinated information flow; instead each handoff network exhibited unique communication patterns and information coordination by two or more influential providers from nursing, medicine, or pharmacy. Most (84%) participants preferred verbal communication. Overall satisfaction with the current communication process varied by unit: 82% of emergency department providers and 54% of the providers working in the admitting units stated they were satisfied or very satisfied. Recommendations for improvement included converting all units to the electronic health record, electronic handoff communication modules and asynchronous multi-professional communication logs. The results of this exploratory study provide a foundation for future research examining how network structure and communication principles can be used to design health information technology that compliments the non-linear information gathering and dissemination behaviors of providers from multiple professions.
    Full-text · Article · Apr 2010 · International Journal of Medical Informatics
  • Source
    • "In this qualitative study of five organizational efforts to reduce disparities in cardiovascular and diabetes care, we identified six key elements that may influence the success of such efforts. These elements can be used by program leaders and researchers to more systematically describe and evaluate the design, implementation and impact of interventions to reduce health care disparities, consistent with calls to examine the relationships between processes and outcomes in quality improvement programs (Gold, Doreian, & Taylor, 2008; Mittman, 2004). The interventions examined in the study were generally broadly focused quality improvement approaches based on the Chronic Care Model (Bodenheimer, Wagner, & Grumbach, 2002), albeit in settings with substantial racial and ethnic minority populations. "
    [Show abstract] [Hide abstract] ABSTRACT: Reducing racial and ethnic disparities in health care has become an important policy goal in the United States and other countries, but evidence to inform interventions to address disparities is limited. The objective of this study was to identify important dimensions of interventions to reduce health care disparities. We used qualitative research methods to examine interventions aimed at improving diabetes and/or cardiovascular care for patients from racial and ethnic minority groups within five health care organizations. We interviewed 36 key informants and conducted a thematic analysis to identify important features of these interventions. Key elements of interventions included two contextual factors (external accountability and alignment of incentives to reduce disparities) and four factors related to the organization or intervention itself (organizational commitment, population health focus, use of data to inform solutions, and a comprehensive approach to quality). Consideration of these elements could improve the design, implementation, and evaluation of future interventions to address racial and ethnic disparities in health care.
    Preview · Article · Nov 2009 · Social Science [?] Medicine
Show more