Dynamics of Innovation Implementation in Social Service Organizations
Peter S. Hovmand, Ph.D.
George Warren Brown School of Social Work
Washington University in St. Louis
Campus Box 1196
One Brookings Drive
St. Louis, MO 63130, USA
(314) 935 7968
phovmand@ wustl.edu
David F. Gillespie, Ph.D.
George Warren Brown School of Social Work
Washington University in St. Louis
Campus Box 1196
One Brookings Drive
St. Louis, MO 63130, USA
(314) 935 6674
davidfg@ fidnet.com
Abstract
Successful implementation of innovations is central to social service organizations’
effectiveness and improvement of services to clients. Yet administrators face a host of challenges
and implementation failures are common. This paper discusses the nature of the innovation
implementation as inherently dynamic, endogenous to the organization, and constrained by
conditions of bounded rationality. Several system dynamics models of innovation
implementation are reviewed from manufacturing, health, and human services in terms of their
appropriateness and evidence base for social services. Recommendations for practice anda
research agenda offered.
Keywords: innovation, implementation, social services, system dynamics
1. Introduction
Current discussions of the barriers to using evidence based practice in social services are
drawing attention to the more general problem of innovation implementation (Proctor, 2004).
While the diffusion literature focuses on decisions to adopt an innovation, implementation
research looks at how new practices get implemented following the decision to adopt (Klein and
Knight, 2005). At least since the work of Zaltman, Duncan, and Holbek (1973), scholars have
pointed to the great difficulty that organizations have with successfully implementing new
practices. Studies of implementation are critical because organizational effectiveness typically
entails the ability to successfully implement new ideas (Aiken and Hage, 1970; Argyris and
Schon, 1978; Borins, 2002).
Innovation Implementation 2
Social service organizations are constantly under pressure to change. The needs of clients
and demands for services change with social conditions, availability and accessibility of services,
and individual client histories. Organizational environments change with federal and state
policies, social attitudes, political climate, professional standards, funding streams, availability of
trained professionals, competitors, and the knowledge base. Organizations also face internal
pressures to change from staff tumover, new management, information systems, budgets,
outcome evaluations, and reorientation of strategic goals. Such pressures motivate organizations
to find, develop and implement new ideas. Initiatives to improve the quality of services typically
involve all three of these domains: client needs, the environment, and internal organizational
processes.
Recent efforts to improve the quality of social services have focused on the
implementation of evidence based practice (EBP). EBP is broadly defined as the “conscientious,
explicit, and judicious use of current best evidence in making decisions about care of individual
patients" (Sackett et al., 1996). EBP is a system for making clinical decisions aimed at
improving the quality of services by maximizing the use of scientifically supported interventions.
The term has now spread to nursing, teaching, and social work, along with controversy that
parallels a drift from its original formulation (Jonson-Reid et al., under review). Despite its
promise, implementation of EBP remains a challenging problem with a call for more studies
addressing organizational barriers (Bartels et al., 2002; Rosen, 1994; Hoagwood et al., 2001;
Newman, Papdopoulous, and Sigsworth, 1998; Rosenheck, 2001; Schoenwald and Hoagwood,
2001).
Addressing barriers to implementing EBP involves understanding the implementation
process as a social phenomenon that is inherently dynamic, endogenous, and constrained by
conditions of bounded rationality. It is dynamic because the focus is on change over time,
endogenous because the process is largely internal to the organization where success builds on
itself and resistance to change is a long-recognized barrier to implementation (Coch and French,
1948), and constrained by conditions of bounded rationality (Kahneman, 2003; Simon, 1979).
That is, implementation of EBP as a social phenomenon is dominated by and therefore best
understood in terms of dynamic complexity.
The primary goal the paper is to characterize the implications of this dynamic complexity
on knowledge development for implementing EBP. The paper is organized as follows. Section
2 describes the epistemological challenge facing organizational and implementation researchers,
and argues for the construction of models of the causal mechanisms underlying implementation.
Section 3 takes up this approach by reviewing several existing models of organizational change
and innovation implementation. And, Section Error! Reference source not found. proposes a
research agenda for advancing the study of implementing EBP.
2. A plea for causal mechanisms in implementation research
We are, in one sense, intimately familiar with the factors contributing to implementation
failures within organizations. Most people can, for example, point to soaring costs and
inadequate budgets, not enough time, high caseloads, employees resisting change, lack luster
performance of managers overseeing the implementation, overly optimistic timelines, disruption
of services, changes in organizational environments, and bad ideas that should never have been
adopted in the first place as reasons for why something did not get implemented. The tendency
is to attempt to fix these factors by increasing budgets, motivating employees, putting more
Innovation Implementation 3
competent managers in charge, investing more time in planning, or setting more realistic
timelines. Y et this focus on fixing particular factors associated with implementation failure is
flawed and inherently misleading because it does not recognize the dynamic interdependencies
that govern the implementation process.
Knowledge of factors contributing to implementation failures or barriers to implementing
EBP is largely based on a set of associations between purported causes and their effects, and
therefore essentially descriptive in nature. For example, most studies of the diffusion and
implementation of evidence based practice refer to factors that increase or decrease adoption or
implementation, but stop short of specifying the causal mechanisms driving change (Hovmand,
Perron, and Proctor, 2005). Those that do mention causal mechanisms tend to either be
qualitative studies or conceptual papers. The lack of well-specified causal mechanisms in
implementation research parallels the general absence of causal mechanisms in other discourses
such as empirically supported psychological treatments (Wampold, 2006; Jensen et al., 2005),
social theory (Hedstrém and Swedberg, 1998), and evaluation research (Brickmayer and Weiss,
2000).
One reason for the dearth of causal explanations is the dilemma that social scientists face
when studying organizations as complex systems within unique environments. Attempts to
develop rigorous universal models come at the price of relevance to the administrator. Schon
(1983) sees this as the dilemma of rigor versus relevance when scientists seek to develop
generalizable knowledge in the tradition of the hard sciences. Instead, recognizing that
organizations are uniquely situated within complex and changing environments, Schon argues
for organizational case studies of prototypical causal patterns or stories. Prototypical causal
patterns are explanations of how the outcomes were achieved. These explanations emphasize
general features of the situation. Such causal stories are not generalizable knowledge in the
sense of giving decision makers universal facts about how one might implement an innovation.
Instead, Schén argues that prototypical causal patterns facilitate what cognitive psychologists
refer to as analogical transfer.
Analogical transfer refers to the application of a solution from one problem to another
problem and is associated with expertise (Novick, 1988; Novick and Hmelo, 1994; Reeves and
Weisberg, 1994). Transfer occurs when a decision maker recognizes surface or structural
similarities between the current problem and a previously solved problem. Novices tend to only
see surface similarities and mimic the solution procedure, while experts are more likely to see
structural similarities between problems with dissimilar surface characteristics and adapt
previous procedures to the current problem. But analogical transfer can either help or hinder
problem solving. When salient features are recognized, positive transfer helps decision makers
find solutions. Negative transfer is also possible. For example, identifying false or irrelevant
similarities - surface or structural - can lead to the wrong strategy for solving the problem. An
emphasis on understanding prototypical causal pattems is motivated by a desire to enable
positive analogical transfers. In the case of implementation research, the type of expertise sought
is an ability to recognize structural similarities in the implementation of innovations across
organizations that differ in their surface characteristics.
Recognizing structural similarities entails being able to compare two or more
representations or mental models of a process. A mental model is a representation of knowledge
about some problem or system that can be manipulated to find a solution (Johnson-Laird, 1983).
For example, in searching for a solution to the problem of implementing evidence based
Innovation Implementation 4
practices, an administrator draws on an abstraction of the organization that can be used to
anticipate the consequences of various actions, and thereby performs a mental simulation of the
organization during the implementation process. Mental models are necessarily simplified and
incomplete, but they also tend to be fuzzy and systematically exclude features leading to
misperceptions of a problem structure and errors in judgement, especially in dynamically
complex environments (e.g., Sterman, 1989, 1989; Moxnes, 2000; Funke, 1991; Brehmer, 1992).
Mental models have therefore long been the targets of interventions where improving the mental
models will lead to better decision making (Axelrod, 1976; Forrester, 1971).
Inferring structural similarities across a set of diverse cases implies a generic structure. A
generic structure is a mental model that represents a structural similarity across a diverse range
of cases. Being able to identify generic structures thus helps decision makers with analogical
transfer and problem solving. We believe that what distinguishes the expert from the novice is
his or her stock of generic structures. Lane (1998) has considered the nature of generic structures
more formally by classifying them into counter-intuitive archetypes, abstracted micro-structures,
and canonical situation models.
Counter-intuitive archetypes refer to qualitative descriptions of causal maps and
dynamics, which are usually accompanied by some type of management principle. Archetypes
are essentially metaphors that facilitate analogical transfer by helping practitioners recognize a
deeper structure and apply the management principle. The difficulty with such archetypes is their
ambiguity. Archetypes are open to different interpretations and hence different management
principles. Moreover, there is heavy reliance on the use of human cognition to draw valid
inferences from such qualitative representations. This is problematic given the nature of mental
models— if we could simulate mental models, then we would not need a method of refining
them.
Abstracted micro-structures represent a different type of generic structure, typically used
as parts in creating more elaborate structures. Micro-structures generate a particular pattern of
behavior such as exponential growth, decline, or oscillation. While they help us recognize an
abstracted structure underlying a particular behavior pattern, they do not necessarily help us
understand the importance of a given micro-structure relative to the others in a particular model.
Micro-structures are generally too limited to represent the entire mental model of a particular
situation.
Canonical situation models hold promise for facilitating analogical transfer. Lane’s
(1998) canonical situation models comes closest to Schén’s (1995) prototypical causal patterns.
Like prototypical causal patterns, canonical situation models represent patterns such as
commodity production cycles, high staff turnover, declining service quality despite investments
in Total Quality Management, and implementation of innovations. For Lane, however, canonical
situation models are more formal mathematical models of causal mechanisms.
What distinguishes a canonical situation model from other formal models is having
passed a family resemblance test. That is, canonical situation models are empirically supported
claims about an underlying structural similarity across a variety cases. What makes a model a
canonical situation model is its passing the family resemblance test (Lane, 2006). As fully
specified mathematical models, canonical situation models provide a means to test a set of
hypotheses about the relationship between model structure and behavior. This allows one to both
refine his or her understanding of the model through the use of mathematical analysis and to test
Innovation Implementation 5
relationships with a wide range of data. The development of such mathematical models provides
a means to accumulate knowledge in way that will maximize the utility to administrators in
social services as they face a series of innovation implementations with evidence based practice.
System dynamics is in a prime position to contribute to this development. Pioneered by
Forrester (1961; Forrester, 1968, 1969, 1971) at the Massachusetts Institute of Technology and
with origins in control systems theory from electrical engineering, system dynamics is a way to
solve problems by understanding how a set of causal feedback mechanisms or feedback loops
interact over time to generate the dynamic behaviors of a system. One of the main advantages of
system dynamics is that it provides a method for developing more precise theories of dynamic
behavior. People often make statements that are so vague and superficial they cannot be proved
wrong. Of course, making precise statements removes ambiguity and opens the possibility of
being wrong. In one sense, any mathematical approach to specifying social theories does this.
What distinguishes system dynamics from other mathematical approaches to model
building is a set of rules for formulating variables and the relationships between them that
enforces a discipline, rigor, and thereby encourages one to develop theories that are internally
consistent in their logic and tested against data. The point of this is not to be right in the sense of
fitting the data, but to have more precise statements that can be communicated and facilitate
learning. This is very much in the spirit of Meehl (1990), where theory becomes more refined
through more precise formulations of causal relationships. Lane makes a similar point when
stressing the opportunities for system dynamics in the agency/structure debate, “This part of
social theory is crying out for a formal yet rich approach to theory building that will allow
connection with empirical data and the elaboration of theories which results in the accumulation
of well-grounded insights” (2001, p. 301). The next section takes up this idea by reviewing
existing system dynamics models of organizational change as potential contributions to theories
for implementing EBP.
3. Model related to implementation of EBP
There are four models from system dynamics literature that are especially relevant to
understanding the dynamics around implementation of evidence based practice: Levin and
Roberts’ (1976) general model of human treatment dropout, Samuel and Jacobsen’s (1997)
model of planned organizational change, Sastry’s model of punctuated organizational change
(1997), and Repenning’s (2002) model of innovation implementation. All four focus on
organizations as the unit of analysis. These models were selected because they reflected some
aspect of the implementation problem for evidence based practice in a social service agency,
used system dynamics, were published in a book or journal article, and cited in the literature.
Although all of the models exist in some form as running simulation models, their
general presentation varies from unsigned causal loop diagrams to stock-and-flow
representations. Presenting one model alone would require simplifying the structure, let alone
four models. So we have chosen to represent overviews of the model in the form of causal-loop
diagrams with polarities assigned to the links and named feedback loops. For some models, this
means that the direction of influence and loop polarities have been worked out from analysis of
the model, while in other cases mechanisms have been named according to descriptions in the
source text. Although this facilitates a comparison of the models with respect to the problem of
implementing evidence based practice, some caution is in order. First, as has been well-noted in
the literature, causal loop diagrams are often misleading in terms of being able to draw accurate
Innovation Implementation 6
inferences about dynamic behavior, and hence the need for simulation. Second, we have
emphasized aspects with respect to a particular problem (implementation of evidence based
practice in social services), and thus our interpretation of these models is necessarily limited.
We have not, for example, tried to discuss all of the implications of these models.
3.1 Levin and Roberts’ model of human service delivery systems
Levin and Roberts (1976) present a general theory of human service delivery systems
where the life cycle of a social service organization is understood in terms of an interaction
between demand for services and services rendered. Their main argument is that client outcomes
are a function of endogenous organizational processes where demand for services is curtailed by
restricting services and lowering quality. A simplified representation of their model is shown in
Figure 1.
Specifically, as the demand for services begins to outstrip the agency’s available
resources for providing high quality services, the agency develops program policies that restrict
services. This has two effects. Restricting services lowers program standards, which lowers
morale, increasing inefficiencies in service delivery, and thereby further reduces available
resources for meeting demand (feedback loop R1 in Figure 1). This process will continue until
the erosion of program policies reduces client satisfaction to the point where demand is low
enough to eliminate the service effectiveness gap (feedback loop B1). Levin and Roberts argue
that more persistent service effectiveness gaps will be addressed through increased community
response, which will lead to greater external funding (B2) and more awareness of the services as
the community mobilizes to increase funding (R2).
Levin and Roberts apply their general theory of human service delivery systems to three
situations: treatment dropout for chemical dependency in a mental health facility, declining
student performance in schools, and shortage dental care supply and demand. Each application
involves a separate simulation model where they explore the dynamics and possible intervention
strategies. All three have relevance to social services. The treatment dropout model, for
example, includes structures representing the efficacy of treatment. The student performance
model includes interactions between teacher, parent, and student expectations. The dental care
model shows how one can model some of the dynamics between latent unmet need for services
and acute services.
The problem of implementation of evidence based practice can be conceptualized as
intervention to reduce the service effectiveness gap by improving program standards. Of
particular interest to researchers focusing on implementation of evidence based practice will be
model structures of how efficacy of empirically supported treatments can deteriorate as a
function of higher demand for services created by improving services and to a lesser extent,
effective advocacy of advocacy groups. If the effectiveness of an empirically supported
treatment is not robust over variations in program policies, then it will be rational for staff to
reject the innovation as ineffective or harmful to clients. These types of effects will be hard to
see because they emerge as the implementation of the innovation leads to more effective services
and thus higher demand, as opposed to being a direct consequence of agency characteristics at
the time of intervention. A social service agency facing this situation would be a victim of its
own success in implementing evidence based practice.
Innovation Implementation 7
Program Standard
Morale Program Policies
a |
quality
+
Service
Effectiveness Gap
Available
Resources . ‘
support
Response
Funds an Pe
+
Political Climate
Figure 1 Causal loop diagram of Levin and Roberts’ general model of human service delivery system, adapted
from Levin and Roberts (1976, p. 36).
Levin and Roberts do not provide empirical support for their reference modes, model
parameters, or equations. Such limitations are probably inherent to work conducted in the mid-
70’s because human service organizations would generally not have been using electronic
databases, making the use of such data an unrealistic expectation for the purpose of their project.
Hence, their models are conceptual. However, most social service organization or programs now
maintain electronic databases of client records and caseloads, which are readily available for
secondary analysis.
3.2 Samuel and Jacobsen’s model of planned organizational change
Samuel and Jacobsen (1997) develop a model of planned organizational change with a
focus on describing how performance changes with the implementation of an innovation. Their
emphasis is on understanding how organizations respond to the gap between desired
performance and current performance. They base their model on the implementation and
organizational change literature, and compare their simulation results against numerical data
from three case studies. Figure 3 shows an overview of their model with the major variables and
feedback mechanisms.
In Samuel and Jacobsen’s model, the initial dip in performance is caused by three
reinforcing mechanisms: costs of implementation (R1 in Figure 3), complexity of change (R2),
and involvement in decision making (R4). Specifically, as more people in the organization get
involved with the implementation process, the costs of lower productivity with using a new
innovation become apparent to the organization and raise questions about the suitability of the
pacing of implementation. Likewise, complexity of the change increases with the number of
people needing to implement it. Both contribute to an increase in resistance to change. And, as
resistance to change increases, fewer employees participate in the decision making process,
Innovation Implementation 8
which leads to even more resistance to change. These three reinforcing mechanisms push
performance even lower and form “vicious cycles” during the initial stage of implementation.
As resistance builds and performance continues to decline, however, inducements or
incentives are used to overcome resistance to change. The increasing level of resistance leads to
the use of performance incentives or reduced caseloads, which have the immediate effect of
lowering resistance to change. As more these inducements are applied, resistance declines and
performance begins to increase, activating a balancing feedback mechanism (B1). At the same
time, inducements add to the overall cost of change. This can lead to a reinforcing mechanism
(R3) where the increased costs lower the suitability of pacing, increase resistance to change, and
further increase the use of inducements.
In Samuel and Jacobsen’s model of organizational change, implementation is a function
of how these five feedback mechanisms interact over time. Innovation implementation will
succeed if the inducements (B1) are sufficient to tip the direction of the three main feedback
mechanisms costs, complexity and involvement (R1, R2, and R4) from being “vicious” cycles
that increase resistance and lower performance to “virtuous” cycles that decrease resistance and
increase performance. Implementation failure can happens when the need for the need for
inducements (R3) begins to dominate.
Samuel and Jacobsen’s model of planned organizational change offers a number of
insights into the problem of implementing evidence based practice. In their model, the
implementation of EBP can be represented as a change to increase performance, where
performance would mean client outcomes. One of the first implications of their model is that
implementation might lead to an initial decrease in client outcomes. The decrease here would be
the result of resistance to change, which would be caused by the complexity of the change (e.g.,
switching over to new forms), costs (e.g., not having enough time to complete paperwork for
monitoring client progress or reimbursement of services), and not participating in the decision
making.
Second, successful implementation of EBP depends on whether or not the inducements
are adequate for tipping the balance of the costs, complexity and involvement feedback
mechanisms. Samuel and Jacobsen’s model assumes that there is no delay between resistance to
change and performance. This is a reasonable assumption when considering activities where the
outcomes are immediately (or almost) observable. For example, over the two-year course of
some implementation effort, accurate completion of billing forms could be observed on a daily,
weekly, or even monthly basis without loss of generality. However, in many social service
interventions, performance can only be observed months or years after the initial intervention.
So the negative aspects of organizational change are immediately observable while the benefits
are only unobservable after long delays. Consequently, it will be harder to get the involvement,
complexity, and costs feedback mechanisms to function as virtuous cycles, and social service
organizations are at greater risk of falling into the trap of becoming addicted to incentives for
reducing resistance to change.
Third, Samuel and Jacobsen’s model of planned organizational change highlights the
importance of workers’ participation in the change process. The complexity and costs associated
with organizational change may be hard to anticipate and control, and generally speaking, most
Innovation Implementation 9
social service agencies have limited resources for offering incentives. However, workers in
social service organizations are more likely to be highly committed to their clients, and therefore
highly invested in change efforts that have the potential to improve the well-being of their
clients. Moving to a more participatory leadership style might therefore be able to change the
direction the direction of the reinforcing effects in the involvement feedback mechanism (R4),
and thus tip the balance of the two other feedback mechanisms. That is, leadership style could
therefore be an important leverage point for managing implementation of evidence based
practice.
£
re)
re) Costs of Change
Size of Target - a
r, eee of
soni R3 need
Initial [ nt for induc
Performance > Performance Suitability of a
2?
Inducements to
Resistance to <= Change
Change induce-
ments
involve-
ment
Participatory Involvement in Exogenous
Leadership Style 4 Decisions Constraints
Figure 2 Causal loop diagram of Samuel and Jacobsen’s model of planned organizational change, adapted from
Samuel and Jacobsen (1997, p. 154).
Samuel and Jacobsen do provide some empirical support for their model by testing their
simulated behavior against numerical reference modes from three separate organizations
undergoing change. In doing so, they demonstrate how their model passes the behavior
reproduction test. Samuel and Jacobsen also draw on some empirical research in formulating
their model and equations. This adds to the empirical support of their model, but it is limited and
falls short of being more rigorously tested. Their model contains a number of relationships that
could be readily tested in a longitudinal study of organizational change. There are now
psychometrically sounds measures of organizational concepts such as resistance to change.
Moreover, longitudinal methods for studying the organizational variables have advanced
considerably in the last 10 years.
Innovation Implementation 10
3.3 Sastry’s model of punctuated organizational change
Sastry’s (1997) develops a model of Tushman and Romanelli’s (1985) theory of
punctuated organizational change. Tushman and Romanelli argue the organizations experience
rapid periods of punctuated change and strategic reorientation to overcome organizational inertia.
Organizational change can then be thought of as altemmating periods of gradual change where
competencies and inertia build. Initially these processes help the organization adjust. However,
as these same competencies and inertia build, it becomes increasingly difficult for the
organization to adjust to a changing environment. A gap between the organization’s strategic
direction and the organization’s environment develops that can only be corrected through a
sudden adjustment in the strategic goals of the organization, which Tushman and Romanelli refer
to as punctuated organizational change. Sastry’s emphasis is on taking the verbal theory from
Tushman and Romanelli’s and testing its logical consistency using system dynamics. Basing the
model on a textual analysis of Tushman and Romanelli and refinement from model testing,
Sastry arrives at the model shown in Figure 3.
For an organization to be effective, it most be both competent in the delivery of services,
and deliver services that reflect the needs and demands of its environment. Sastry’s model
represents this by having organizational performance as determinants of (1) appropriateness of
the organization's strategic orientation to its environment, and (2) competence (Figure 3). The
basic argument is that initially when organizations start out to meet the needs of the environment,
their strategic orientation is generally appropriate but because they are new, they do not have
much competence in the delivery of services. So organizations build up competencies that lead
to improvements in performance (R2 in Figure 3), while leaning and socialization
institutionalizes the means of service delivery (R1 in Figure 3).
However, as the organization’s environment gradually changes, the appropriateness of
the organization’s strategic orientation declines, leading to a decrease in performance.
Unfortunately, the same mechanisms that leads to increasing performance early on now makes it
difficult for the organization to realign its strategic goals. Institutionalization that contributes to
high performance also contributes to increasing inertia and erodes the organization's ability to
change its orientation (R3 in Figure 3). So the organization's strategic appropriateness continues
to decline until there is a strategic reorientation (B1 in Figure 3).
There are a number of ways to use Sastry’s model to understand the challenges of
implementing evidence based practice. First, one can consider the initial decision of an
established agency’s board of directors or executive direct to commit the organization to a
particular model of using evidence based practice. In this situation, client outcomes and
organizational performance can oscillate between high and low organizational effectiveness.
Consider, for example, the initial decision to implement evidence based practice within
an organization. In the initial stages after the strategic reorientation, competence and
organizational performance are low. Client outcomes begin to improve as the staff gain
experience and procedures are institutionalized. While the strategic orientation of the agency
and environment are aligned, improvements in competence lead to better client outcomes and
organizational performance. Eventually, however, the model of implementing evidence based
practices obsolesces, and organizational performance begins to decline. However,
Innovation Implementation 11
institutionalization has eroded the agency’s ability to change, and this makes it increasingly
difficult for the agency to reorient its services and adopt new models of implementing evidence
based practices. Consequently, client outcomes and organizational performance decline further.
Eventually the organization undertakes a new strategic reorientation, starting a new period of
organizational change. In this scenario, client outcomes and organizational performance
oscillate.
Strategic
Orientation .
Required Appropriateness
Strategic aia
Orientation Bi
change +
in orien-
+ tation
Change in Strategic _ + Pressure to if Perceived +
ms Change Performance Boe me
(cy s
(cy
etence Competence
See OS te
RL
inertia
tai)
tai) to Leaming,
change Socal,
Ability to Change
Figure 3 Causal loop diagram of Sastry’s model of punctuated organizational change, adapted from Sastry (1997, p.
244).
A key implication from this is that the more one seeks to institutionalize processes to
maximize implementation of evidence based practice, the more likely one is to also limit the
organization’s ability to adapt to a changing environment. Thus short-term performance gains
might be offset or limited by longer-term performance declines.
A second aspect of Sastry’s model concems startup agencies. Startups do not have
inertia. They therefore have a greater ability to both orient their strategic goals to the needs of
the community and implement evidence based practice. Once the startup develops competence,
the better environmental fit means that the startup could potentially outperform more established
organizations. Such startups would be able to address the idiosyncratic needs of the community
and implement evidence based practice. A successful startup would therefore gain recognition in
the community as an innovator.
The tendency would be to attribute the startups success to leadership, an innovative
business model, etc. This could bring more pressure on established organizations to adjust their
strategic orientation and adopt new models of delivering services. High inertia in established
Innovation Implementation 12
organizations would slow this process, and organizational performance would continue to
decline while the startup looked even more successful.
The startup’s advantages are, however, temporary. As the agency institutionalizes its
procedures and service delivery model, the agency develops inertia that improves competence
and outcomes. At the same time, building inertia erodes the organization's ability to change.
Paradoxically, the more confident the leadership is with their service delivery model, the more
vulnerable the board and executive director are to attribute the initial success to the plan as
opposed to initial conditions of starting the organization. If the agency does not invest in a
longer term strategy, then it risks become a mediocre organization resistant to change; arguably
type of organization that created the niche for innovation.
There are several lessons here about successfully implementing evidence based practice.
First, startup agencies with their low inertia may be easier sites to implement evidence based
practice than established social service agencies. Second, the agency’s capacity to transition
from initial startup into a more mature and established organization is critical to the successful
implementation of evidence based practice over the long-term.
Sastry’s model is primarily aimed at testing the causal logic within existing theory.
Testing consists of comparing the model’s behavior against descriptions of organizational
behavior from Tushman and Romanelli (1985). Although there is empirical support for
organizational change being punctuated, the model itself is not tested against empirical data.
Sastry’s question is tightly focused on whether or not Tushman and Romanelli’s theory is
logically consistent, and not on whether or not the theory agrees with data. The testing informs
refinements in the model and theory, but is not empirically based in the sense that the causal
relationships have been rigorously testing against experimental or quasi-experimental time-series
designs.
3.4 Repenning’s (2002) model of innovation implementation
Repenning (2002) constructs a model of the process of innovation implementation to
understand the paradox in organizational theory where organizations reject innovations that
improve effectiveness when successfully implemented. Repenning gives the example of Total
Quality Management (TQM), an innovation that improves effectiveness and yet fails to be
implemented. TQM has been shown to improve the productivity, quality, and market
competitiveness of firms, yet successful implementation is rare. The issue here is not about the
translation of research to practice. Rather, it is a question about why interventions that by all
measures improve effectiveness are not implemented.
Repenning’s approach is similar to Sastry in using simulation as a tool for developing
more internally consistent theories of an organizational phenomenon. Repenning first develops
causal loop diagrams to represent three processes implicated in implementation: reinforcement,
diffusion of innovation, and normative pressures. These three processes are represented as in
Figure 4. The variables and linkages in the diagram are generally well accepted and empirically
based.
Innovation Implementation 13
In the process of reinforcement, the more committed employees are to the innovation, the
more effort they apply to using the innovation and the better the results, which feeds back to
increase commitment to the innovation (feedback loop R1 in Figure 4). Reinforcement can
function as either a virtuous or vicious cycle. For example, if commitment drops, effort declines
along with results, and this justifies a further decline of commitment. In the diffusion of
innovation, results (positive or negative) are passed on to others and influence overall
commitment, which drives the effort allocated to the innovation, and hence results (feedback
loop R2 in Figure 4). This is also a reinforcing mechanism. Good word of mouth will work as a
virtuous cycle, while bad word of mouth will work as a vicious cycle. Lastly, there is the
managers’ process of applying normative pressures. The larger the commitment gap between the
commitment to the innovation and the managers’ goal, the more pressure managers apply to
increasing commitment, which reduces the commitment gap (feedback loop B1 in Figure 4).
This works as balancing loop and depends on the managers’ goal for commitment.
Observation of
Effort- Results
Linkage by Others
R2 q
ee, of Magagers' Goal for
diffusion Cc. r nt
(aX 5 Commitment G; c,
Commitment to
Results the Innovation
Relator
coment, ‘So Ne
Effort Normative
Allocated to Pressure from.
the Innovation Managers
Figure 4 Causal loop diagram of Repenning’s model of implementation of TQM, adapted from Repenning’s (2002,
p. 112).
Next, Repenning develops a simulation model to analyze the implications of these three
feedback mechanisms to try and resolve the paradox of how innovations that should improve
effectiveness are rejected. Repenning finds that these three feedback mechanisms are able to
explain both successful and failed implementation of an innovation. He also finds that duration
of managers’ goal for commitment play a key role in distinguishing successful from unsuccessful
implementation. Specifically, the relationship between how long managers apply normative
pressures and successful implementation is highly nonlinear. In essence, there is a “tipping
point” where the implementation process changes from being management driven to endogenous
to the work process. Past this tipping point, management can withdraw support and the
implementation continues. Before this tipping point, withdrawal of management support will
Innovation Implementation 14
result in implementation failure. Differences of as little as one month can make the difference
between a successful and failed implementation.
Repenning’s model is grounded in the empirical literature on organizational theory
around well accepted mechanisms. The model is also able to explain the two widely observed
outcomes of an implementation process. While the model is grounded on empirically based
organizational theory, the model itself has not been subjected to empirical studies. That is, the
model contains a number of hypotheses about the conditional relationships between variables
and temporal relationships between dominant mechanisms.
Repenning’s model has two important implications for the implementation of evidence
based practice. First, implementation failures of evidence based practice are commonly assumed
to derive from the innovation being ineffective outside the research setting. Repenning’s work
shows how this need not be the case, and that the determinants of fidelity may trump questions
about efficacy and effectiveness.
Second, the duration of managers’ support plays a key role in the outcome, not just the
initial conditions. That is, in Repenning’s model what distinguishes the successful from failed
implementation is not their readiness to change. Both trajectories start out with the same initial
conditions and in this sense are both equally ready to change. What distinguishes the failed from
successful implementation is whether or not managers remained committed beyond a tipping
point. One could argue that this is a dimension of an organizations’ readiness to change, for
example, by including an indicator reflecting the likelihood that managers’ will remain
committed to the change process. However, this seems to largely ignore how determinants of
managers’ commitment are endogenous to change process itself. Managers respond to the state
of the organization, make an assessment, and then decide on a course of action. This is what
managers do.
Third, given that the duration of managers’ support plays such a key role, increasing the
strength and accessibility of the evidence behind the innovation could be a key leverage point for
successful implementation of evidence based practice. If extending managers’ commitment to
implementing an evidence based practice by as little as one month might make the difference
between a successful and failed implementation, then increasing the availability of evidence and
confidence of managers’ in the effectiveness of the intervention could be a key determinant of
success. This is an area ripe for future empirical research with potentially high impact on
improving the effectiveness of social service organizations.
4. Discussion
All four models help advance the development of causal models of innovation
implementation. By identifying a number of causal feedback mechanisms and representing them
as system dynamics models, assumptions and relationships between key concepts are made
explicit and testable. As running simulation models, they provide a transparent means of
evaluating claims about the relationships between the feedback mechanisms and dynamic
behavior of the system, leading to a stronger, more internally consistent theory of organizational
change. Although all four models use causal loop diagrams to represent the verbal theory,
simulation is an essential tool for overcoming the cognitive limitations of trying to draw valid
inferences about the causal relationships in dynamic nonlinear systems. Moreover, the models
Innovation Implementation 15
can be extended and adapted to new situations. This opens up the possibility of developing
innovation implementation models specific to evidence based practice in social service
organizations.
The four models differ somewhat in their time horizons. Samuel and Jacobsen, and
Repenning’s models focus on a short timescale around a single innovation implementation.
Sastry considers the organization over a longer period of where there is sufficient time for a gap
between organization’s strategic goals and its environment to emerge, while Levin and Robert's
consider the complete life cycle of a human service delivery system. Differences in time
horizons imply differences in the problems being considered. The salience of the implications
for managers and researchers will therefore depend on what period of time is of most interest.
Models focusing on shorter time periods will be most relevant to the questions around a single
implementation of evidence based practice, while models focusing longer time horizons will
have more relevance to questions around strategic planning and sustaining long-term growth.
The existing models of organizational change and innovation implementation can help us
better conceptualize the problem of implementing EBP in social service organizations by
providing more logically consistent social theories. They can also help us recognize structural
similarities between seemingly dissimilar implementation problems. They do not provide
empirically based advice or guidelines on how to proceed with implementing evidence base
practice within a specific organization. Instead, the models reviewed here provide prototypical
causal stories of how an implementation process might unfold, the causes of failure in each, and
what one might do to increase the likelihood of successful implementation.
More research is needed to develop a model of implementing evidence based practice,
along with more rigorous empirical studies testing the claims implicit within the models against
actual organizational behavior. Toward this end, we conclude by prosing the following research
activities as important steps toward the development of a generic structure of the problem facing
administrators in social service agencies with implementation of evidence based practice.
1. Replicate system dynamics studies of implementation against current data. One of the
main benefits of system dynamics in having explicit formulations of causal relationships
is that one can replicate previous simulation studies. This would involve building,
calibrating, and testing existing models such as those described here against
organizational data from implementation efforts. The results would deepen our
understanding of existing models and where they fail when tested against data.
2. Evaluate the impact of delays in client outcomes on implementation. To a large extent,
all the models assumed that performance data was immediately available. This is
generally not the case with social service organizations, and introducing such delays
between program activities and clinical outcomes could introduce a new dynamics that
would be important to consider for the effective delivery of social services. These need
to be studied both theoretically and empirically as a way to develop more confidence in
the application of system dynamics modeling to implementation research.
3. Test resource allocation policies against actual behavior. All of these models contain
rate expressions and table functions that have, for the most part, not been tested against
actual behavior of service providers in social service organizations. In particular,
questions arise about how providers balance adhering to the fidelity of their interventions
Innovation Implementation 16
against resources (time, billable hours) and large caseloads. Research on organizational
culture in social service organizations can have a major effect on outcomes, even when
controlling for large caseloads. This suggests that how workers allocate their resources to
ensuring treatment fidelity might vary by the organization and its culture. More research
is needed to better understand how these types of expressions vary by the organization.
4. Document stakeholders’ mental models of implementing EBP. Critical to the
development of a model of implementing EBP would be a need to establish some face
validity of the concepts and perceptions of the problem with implementing EBP. Of
particular interest will be descriptions of the mental models that administrators, clinical
supervisors, and direct service providers draw on when considering the problem of
innovation implementation. More qualitative research is needed to establish the existence
of the innovation implementation issues facing administrators within social service
organizations.
5. Build system dynamics models of implementing EBP within a social service organization.
The implications discussed within this paper should be studied by building and
simulating a model of the implementation process under various conditions and agencies.
This would provide a way to assess to what extent implementing EBP differs from
implementing other types of innovations, as well as help identify to what extent one or
more of the existing models could be considered a canonical situation model for
innovation implementation. More system dynamics modeling is needed to understand the
impact of differences characteristics such as the type of innovation, type of organization,
and organizational culture on the dynamics of innovation implementation.
6. Conduct prospective or evaluation research of innovation implementation using system
dynamics. None of these models considered the models of innovation implementation
within a prospective study. If the goal of developing generic or canonical situation
models is to develop knowledge that helps administrators make better decisions as they
plan the implementation of a new innovation, then it seems critical to understand how the
use of system dynamics models contributes to outcomes. More prospective quasi-
experimental and evaluation research on the impact of stakeholders’ using system
dynamics on managing organizational change is needed in order to assess and improve
the utility of system dynamics research for problem solving and organizational leaming.
7. Develop and empirically test an intervention to increase the duration of managers’
commitment to implementing an evidence based practice. Repenning’s work suggests
that the duration of managers’ commitment to implementing an innovation may be a key
leverage point. Developing and testing interventions that increase the duration of
managers’ commitment by as little one month seems entirely feasible. Moreover, this
could involve any number of strategies, from improving the attitudes toward evidence
based practice to making the evidence behind innovations more accessible to managers.
This would be an innovative line of research, in part, because much of the dissemination
of innovations research has focused on making evidence based practices more accessible
to the clinicians, as opposed to the managers as key decision makers.
Innovation Implementation 17
References
Aiken, and Hage. 1970.
Argyris, C., and Donald A. Schon. 1978. Organizational learning: A theory of action
perspective. Menlo Park, CA: Addison-Wesley Publishing.
Axelrod, R., ed. 1976. Structure of decision: The cognitive maps of political elites. Princeton,
NJ: Princeton: Princeton University Press.
Bartels, Stephen J., Aricca R. Dums, Thomas E. Oxman, Lon S. Schneider, Patricia A. Arean,
George S. Alexopoulos, and Dilip V. Jeste. 2002. Evidence-Based Practices in Geriatric
Mental Health Care. Psychiatric Services 53 (11):1419-1431.
Borins. 2002.
Brehmer, B. 1992. Dynamic decision making: Human control of complex systems. Acta
Psychologica 81:211-241.
Brickmayer, J. D., and C. H. Weiss. 2000. Theory-based evaluations in practice: What do we
learn? Evaluation Review 24 (4):407-431.
Coch, and French. 1948.
Forrester, J. W. 1961. Industrial dynamics. Waltham: Pegasus Communications, Inc.
———. 1968. Market growth as influenced by capital investment. Industrial Management
Review 9:83-105.
———. 1969. Urban dynamics. Cambridge, MA: MIT Press.
——-—.1971. Principles of systems. Waltham: Pegasus Communications, Inc.
Funke, J. 1991. Solving complex problems: exploration and control of complex systems. In
Complex problem solving: Principles and mechanisms, edited by R. Stemberg and P.
Frensch. Hillsdale, NJ: Lawrence Erlbaum.
Hedstrom, P., and R. Swedberg, eds. 1998. Social mechanisms: An analytical approach to social
theory. New Y ork: Cambridge University Press.
Hoagwood, Kimberly, Barbara J. Burns, Laurel Kiser, Heather Ringeisen, and Sonja K.
Schoenwald. 2001. Evidence-Based Practice in Child and A dolescent Mental Health
Services. Psychiatric Services 52 (9):1179-1189.
Hovmand, Peter, Brian Perron, and Enola Proctor. 2005. Getting evidence based practice to work
within organizations: A system dynamics based meta-analysis. Paper read at Society for
Social Work Research, at Miami, FL.
Jensen, P. S., R. Weersing, K. E. HOagwood, and E. Goldman. 2005. What is the evidence for
evidence-based treatments? A hard look at our soft underbelly. Mental Health Services
Research 7 (1):53-74.
Johnson-Laird, P. 1983. Mental models: Towards a cognitive science of language, inference and
consciousness. Cambridge, MA: Harvard University Press.
Jonson-Reid, M., B. Drake, Peter S. Hovmand, and Luis H. Zayas. under review. Correcting the
drift in evidence based practice. J ournal of Social Work Education.
Kahneman, Daniel. 2003. Maps of bounded rationality. The American Economic Review 93
(5):1449-1475.
Klein, Katherine J., and Adrew P. Knight. 2005. Innovation implementation: Overcoming the
challenge. Current Directions in Psychological Science 14 (5):243-246.
Lane, David C. 1998. Can we have confidence in generic structures? J ournal of Operational
Research Society 49:936-947.
Innovation Implementation 18
———. 2001. Rerum cognoscere causas: Part II--Opportunities generated by the
agency/structure debate and suggestions for clarifying the social theoretic position of
system dynamics. System Dynamics Review 17 (4):293-309.
———. 2006. Personal communication, February 21.
Levin, Gilbert, and Edward B. Roberts. 1976. The dynamics of human service delivery.
Cambridge, MA: Ballinger Publishing Company.
Meehl, Paul E. 1990. Appraising and amending theories: The strategy of Lakatosian defense and
two principles that warrant it. Psychological Inquiry 1 (2):108-141.
Moxnes, E. 2000. Not only the tragedy of the commons: misperceptions of feedback and policies
for sustainable development. System Dynamics Review 16 (4):325-348.
Newman, M., I. Papdopoulous, and J. Sigsworth. 1998. Barriers to evidence-based practice.
Intensive and Critical Care Nursing 14 (5):231-238.
Novick, Laura R. 1988. Analogical transfer, problem similarity, and expertise. Journal of
Experimental Psychology 14 (3):510-520.
Novick, Laura R. , and Cindy E. Hmelo. 1994. Transferring symbolic representations across
nonisomorphic problems. Journal of Experimental Psychology: Learning, Memory, and
Cognition 20 (6):1296-1321.
Proctor, Enola. 2004. Leverage points for the implementation of evidence-based practice. Brief
Treatment and Crisis Intervention 4 (3):227-242.
Reeves, Lauretta M., and Robert W. Weisberg. 1994. The role of content and abstract
information in analogical transfer. Psychological Bulletin 115 (3):381-400.
Repenning, N. P. 2002. A simulation-based approach to understanding the dynamics of
innovation implementation. Organization Science 13 (2):109-127.
Rosen, A. 1994. Knowledge use in direct practice. Social Service Review 68 (4):561-577.
Rosenheck, Robert A. 2001. Organizational Process: A Missing Link Between Research and
Practice. Psychiatric Services 52 (12):1607-1612.
Sackett, D. L., W. M. C. Rosenberg, J. A. M Gray, R. B. Haynes, and W. D. Richardson. 1996.
Evidence based medicine: what it is and what it isn't. British Medical J ournal 312 (71-
72).
Samuel, Y itzhak, and Chanoch Jacobsen. 1997. A system dynamics model of planned
organizational change. Computational and mathematical organization theory 3 (3):151-
171.
Sastry, M. Anjali. 1997. Problems and paradoxes in a model of punctuated organizational
change. Administrative Science Quarterly 42 (2):237-275.
Schoenwald, Sonja K., and Kimberly Hoagwood. 2001. Effectiveness, transportability, and
dissemination of interventions: What matters when? Psychiatric Services 52 (9):1190-
1197.
Schon, Donald A. 1983. The reflective practitioner: How professionals think in action. New
Y ork, NY: Basic Books.
———. 1995. The study of organizations. In Rethinking knowledge: Reflections across
disciplines, edited by R. F. Goodmand and W. R. Fisher. Albany, NY: SUNY Press.
Simon, Herbert. 1979. Rational decision making in business organizations. The American
Economic Review 69 (4):493-513.
Sterman, J. D. 1989. Misperceptions of feedback in dynamic decision making. Organizational
Behavior and Human Decision Processes 43 (3):301-335.
Innovation Implementation 19
———. 1989. Modeling managerial behavior: Misperceptions of feedback in a dynamic decision
making experiment. Management Science 35 (3):321-339.
Tushman, Michael L., and Elaine Romanelli. 1985. Organizational evoluation: A metamorphsis
and model of convergence and reorientation. In Research in Organizational Behavior,
edited by L. L. Cummings and B. M. Staw. Greenwich, CT: JAI Press.
Wampold, B. E. 2006. Nota scintilla of evidence to support empirically supported treatments as
more effective than other treatments. In Evidence-based practices in mental health:
Debate and dialogue on the fundamental quesitons, edited by J. C. Norcorss, L. E.
Beutler and R. F. Levant. Washington, DC: American Psychological Association.
Zaltman, Duncan, and Holbek. 1973.
