Diffusion Of Innovations In Service Organizations: Systematic Review And Recommendations

Abstract

This article summarizes an extensive literature review addressing the question, How can we spread and sustain innovations in health service delivery and organization? It considers both content (defining and measuring the diffusion of innovation in organizations) and process (reviewing the literature in a systematic and reproducible way). This article discusses (1) a parsimonious and evidence-based model for considering the diffusion of innovations in health service organizations, (2) clear knowledge gaps where further research should be focused, and (3) a robust and transferable methodology for systematically reviewing health service policy and management. Both the model and the method should be tested more widely in a range of contexts.

Full text

Diffusion of Innovations in Service
Organizations: Systematic Review
and Recommendations
TRISHA GREENHALGH, GLENN ROBERT,
FRASER MACFARLANE∗,PAULBATE,
and OLIVIA KYRIAKIDOU∗
University College London; ∗
University of Surrey
This article summarizes an extensive literature review addressing the question,
How can we spread and sustain innovations in health service delivery and or-
ganization? It considers both content (defining and measuring the diffusion
of innovation in organizations) and process (reviewing the literature in a sys-
tematic and reproducible way). This article discusses (1) a parsimonious and
evidence-based model for considering the diffusion of innovations in health
service organizations, (2) clear knowledge gaps where further research should
be focused, and (3) a robust and transferable methodology for systematically re-
viewing health service policy and management. Both the model and the method
should be tested more widely in a range of contexts.
Key Words: Diffusion of innovation, systematic review, implementation.
This article summarizes the findings of a
systematic literature review of the diffusion of service inno-
vations. The United Kingdom Department of Health explic-
itly commissioned this work, which was carried out between October
2002 and December 2003, for its National Health Service’s exten-
sive modernization agenda (UK Department of Health 2001). Our
review, which supplements and extends previous overviews and meta-
analyses (Damanpour 1991, 1992, 1996; Granados et al. 1997; Meyers,
Sivakumar, and Nakata 1999; Rogers 1995; Tornatsky and Klein 1982;
Address correspondence to: Trisha Greenhalgh, University College London, Room
403, Holborn Union Building, Highgate Hill, London N19 5LW, United
Kingdom (e-mail: p.greenhalgh@pcps.ucl.ac.uk).
The Milbank Quarterly, Vol. 82, No. 4, 2004 (pp. 581–629)
c
2004 Milbank Memorial Fund. Published by Blackwell Publishing.
581

582 Trisha Greenhalgh et al.
Wejnert 2002; Wolfe 1994), focuses primarily, but not exclusively, on
research studies of health care. Because of the size and scope of our re-
view, we cannot describe all our findings or discuss all our sources in this
article. Instead, we encourage interested readers to read the complete
project report (Greenhalgh et al. 2005a).
We defined a systematic review as a review of the literature according to
an explicit, rigorous, and transparent methodology. We defined innova-
tion in service delivery and organization as a novel set of behaviors, routines,
and ways of working that are directed at improving health outcomes,
administrative eff iciency, cost effectiveness, or users’ experience and that
are implemented by planned and coordinated actions. We distinguished
among diffusion (passive spread), dissemination (active and planned ef forts
to persuade target groups to adopt an innovation), implementation (active
and planned efforts to mainstream an innovation within an organiza-
tion), and sustainability (making an innovation routine until it reaches
obsolescence). But we did note an ambiguity in the notion of sustain-
ability (i.e., the longer an innovation is sustained, the less likely the
organization will be open to additional innovations).
The breakdown of sources that contributed to the final report is
shown in Figure 1. Because formal search techniques (e.g., entering
32 journals
Hand search
6,000 titles/abstracts
15 databases
Electronic search
105 books
Library search
1,024 full text papers
and book chapters
appraised
495 sources in final report
282 non-
empirical
213 empirical
studies
166 papers
References
of references
Citation
tracking
figure 1. Summary of Sources Contributing to the Systematic Review

Innovation in Service Organizations 583
index terms or key words in electronic databases) drew a poor yield, we
relied mainly on “snowball” methods (pursuing references of references
and using citation-tracking software) and sought advice on sources from
experts in various fields. Our search strategy was designed to concen-
trate on the service sector, particularly health care. Our original inclusion
criteria were (1) studies from the health care sector; (2) those that had
addressed innovation in service delivery and organization; (3) those that
had looked specifically at the diffusion, dissemination, implementation,
and/or routinization of these innovations; and (4) those that met our
stringent criteria for methodological quality. But as the review unfolded,
two things became clear: first, in many areas, the evidence meeting all
these criteria was sparse, and second, we could gain critical insights from
beyond the parameters we had set. We therefore extended our criteria
to a wider range of literature. In particular, we added both overview
articles and “landmark” empirical studies from outside the health sec-
tor if they had important methodological or theoretical lessons for our
research question. The literature on the sustainability of service innova-
tions was, incidentally, very sparse, and so we did not include it in this
article.
To help explore this large and heterogeneous literature, we developed
a new technique, which we called meta-narrative review.Itissummarized
in Box 1 and explained in detail in a separate paper (Greenhalgh et al.
2005b). A meta-narrative is the unfolding “storyline” of research in a
particular scientific tradition (defined as a coherent body of theoretical
knowledge and a linked set of primary studies in which successive studies
are influenced by the findings of previous studies; see Kuhn 1962). We
mapped the meta-narratives (i.e., we traced the historical development of
concepts, theory, and methods in each research tradition) by identifying
the seminal theoretical and overview papers and books and analyzing
the conceptual and theoretical models proposed by recognized experts
in each field.
BOX 1
Phases in Meta-Narrative Review
1. Planning Phase
a. Assemble a multidisciplinary research team whose back-
ground encompasses the relevant research traditions (an

584 Trisha Greenhalgh et al.
initial scoping phase may be needed before the definitive
research team is appointed).
b. Outline the initial research question in a broad, open-ended
format.
c. Define outputs in collaboration with funder or client.
d. Set up a series of regular, face-to-face review meetings, in-
cluding planned input from external peers drawn from the
intended audience for the review.
2. Search Phase
a. Lead the initial search by intuition, informal networking,
and “browsing” in order to map the diversity of perspectives
and approaches.
b. Search for seminal conceptual papers in each research
tradition by tracking references of references. Evaluate
these by the generic criteria of scholarship, comprehen-
siveness, and contribution to subsequent work within the
tradition.
c. Search for empirical papers by electronically searching key
databases, hand-searching key journals, and “snowballing”
(references of references or electronic citation tracking).
3. Mapping Phase
Identify (separately for each research tradition):
a. The key elements of the research paradigm (conceptual,
theoretical, methodological, and instrumental).
b. The key actors and events in the unfolding of the tradition
(including the main findings and how they were discovered).
c. The prevailing language and imagery used by scientists to
“tell the story” of their work.
4. Appraisal Phase
Using appropriate critical appraisal techniques:
a. Evaluate each primary study for its validity and relevance to
the review question.
b. Extract and collate the key results, grouping together com-
parable studies.
5. Synthesis Phase
a. Identify all the key dimensions of the problem that have
been researched.

Innovation in Service Organizations 585
b. For each dimension, give a narrative account of the contri-
bution (if any) by each separate research tradition.
c. Treat conflicting findings as higher-order data, and ex-
plain them in terms of contestation among the different
paradigms from which the data were generated.
6. Recommendations Phase
Through reflection, multidisciplinary dialogue, and consultation
with the intended users of the review:
a. Summarize the overall messages from the research literature
along with other relevant evidence (budget, policymaking
priorities, competing or aligning initiatives).
b. Distill and discuss recommendations for practice, policy,
and further research.
Source: Greenhalgh et al. 2005b.
For each empirical study, we developed a data extraction form (avail-
able on request) to summarize the research question, study design, va-
lidity and robustness of methods, sample size and power, nature and
strength of findings, and validity of conclusions. We modified published
critical appraisal checklists to assess the quality of primary studies eval-
uating service interventions, qualitative research, mixed-methodology
case studies, action research, and process. We then divided the pri-
mary studies’ findings into six broad categories: (1) the innovation itself;
(2) the adoption/assimilation process; (3) communication and influence
(diffusion and dissemination, including social networks, opinion leader-
ship, champions, and change agents); (4) the inner (organizational) con-
text, including both antecedents for innovation in general and readiness
for particular innovations; (5) the outer (interorganizational) context,
including the impact of environmental variables, policy incentives and
mandates, and interorganizational norms and networking; and (6) the
implementation process. Within each category, we identified subtopics
and painted a rich picture of each by grouping together the contributions
from different research traditions.
Because different researchers in different traditions generally concep-
tualized their topic differently; used different language and metaphors
for diffusion, dissemination, and implementation; asked different ques-
tions; privileged different methods; and used different criteria to judge
“quality” and “success,” we used narrative, rather than statistical, syn-
thesis techniques (Dixon-Woods et al. 2004). We highlighted the

586 Trisha Greenhalgh et al.
similarities and differences of the findings from different research tra-
ditions and considered the reasons for the differences. In this way, the
heterogeneity of approaches and “contradictions” in findings could be
turned into data and analyzed systematically. Based on the evidence from
the primary studies, we developed a unifying conceptual model (Figure 3)
and tested the model on four case studies (telemedicine, integrated care
pathways, general practitioner fund holding, and the UK’s electronic pa-
tient record), which are analyzed in detail in the full report (Greenhalgh
et al. 2005a).
We graded the overall evidence supporting each of our conclusions
using a modified version of the World Health Organization Health
Evidence Network (WHO-HEN) criteria (Øvretveit 2003):
•Strong direct evidence: consistent findings in two or more empirical
studies of appropriate design and high scientific quality undertaken
in health service organizations.
•Strong indirect evidence: consistent findings in two or more empirical
studies of appropriate design and high scientific quality, but not
from a health service organization.
•Moderate direct evidence: consistent findings in two or more empirical
studies of less appropriate design and/or of acceptable scientific
quality undertaken in health service organizations.
•Moderate indirect evidence: consistent findings in two or more empir-
ical studies of less appropriate design and/or of acceptable scientific
quality, but not from health service organizations.
•Limited evidence: only one study of appropriate design and acceptable
quality available, or inconsistent findings in several studies.
•No evidence: no relevant study of acceptable scientific quality
available.
Key Research Areas
Because of the very large number of empirical sources identified, we
have cited in this article only illustrative studies and/or overviews, and
we have not given citations for statements for which we had only limited
evidence. For a full list of primary sources, please see our main report
(Greenhalgh et al. 2005a). We identified 13 research areas that had,
largely independently of one another, provided evidence relevant to the
diffusion of innovations in health service organizations (Table 1). Four
of these traditions can be classified as “early diffusion research”:

Innovation in Service Organizations 587
TABLE 1
Research Traditions Relevant to Diffusion of Innovations in Health Service Organizations
“Diffusion of Innovations”
Research Tradition Academic Discipline Definition and Scope Conceptualized as
1. Rural sociology Sociology Study of rural society and the relationships among
its members, especially the influence of social
structures and norms on behaviors and practices.
Influence of social norms and values on
adoption decisions; networks of social
influence.
2. Medical sociology Sociology As above for medical society. As above. Specifically, the norms,
relationships, and shared values that
drive clinician behavior (e.g., adoption
of guidelines).
3. Communication
studies
Psychology Study of human communication, including both
interpersonal and mass media.
Structure and operation of
communication channels and
networks. Interpersonal influence (e.g.,
impact of “experts” versus “peers” on
decision making).
4. Marketing Interdisciplinary
(psychology and
economics)
Study of the production, distribution, and
consumption of goods and services.
Affordability, profitability, discretionary
income, market penetration, media
advertising, supply, and demand.
5. Development
studies
Interdisciplinary
(anthropology, sociology,
economics, political
science, information and
communications
technology)
Study of the adoption, adaptation, and use of
technology, especially in development.
Barriers to the use of more advanced
technologies (e.g., labor-saving
machinery, computers).
6. Health promotion Interdisciplinary (social
psychology,
epidemiology,
marketing)
Study of strategies and practices to improve the
health and well-being of populations (draws on
and overlaps with communication studies).
“Reach” and “uptake” of positive lifestyle
choices in populations targeted by
health promotion campaigns.

588 Trisha Greenhalgh et al.
TABLE 1—Continued
“Diffusion of Innovations”
Research Tradition Academic Discipline Definition and Scope Conceptualized as
7. Evidence-based
medicine
Clinical epidemiology Study of the spread of best (research) evidence
on managing diseases and symptoms.
Filling a “knowledge gap” or “behavior
gap” in targeted clinicians.
8. Structural
determinants of
organizational
“innovativeness”
Organization and
management
Study of how an organization’s structure
influences its function in relation to the use
of new ideas and practices.
Organizational attributes influencing
“innovativeness,” like size, slack
resources, and hierarchical versus
decentralized lines of management.
9. Studies of
organizational
process, context,
and culture
Interdisciplinary
(organization and
management, sociology,
anthropology)
Study of the development and impact of culture
(meaning systems, language, traditions,
accepted ways of doing things) in
organizations and professional groups.
Changes in culture, values, and identities.
10. Interorganizational
studies (networks
and influence)
Interdisciplinary
(organization and
management, sociology)
Study of interorganizational norms, fashions,
and influence.
Interorganizational fads and fashions,
spread through social networks.
11. Knowledge
utilization
Interdisciplinary
(organization and
management, information
and communications
technology, sociology)
Study of how individuals and teams acquire,
construct, synthesize, share, and apply
knowledge.
Transfer of knowledge, both explicit
(formal and codified, as in a guideline)
and tacit (informal and embodied, as in
“knowing the ropes”).
12. Narrative studies Interdisciplinary (literature,
sociology, anthropology)
Study of stories (here, those told in and about
organizations). Use of storytelling as a tool
for dissemination and change in
organizations.
The telling, retelling, and interpretation
of stories. Innovators as characters
(heroes, underdogs) in a story of
change. Innovation as social drama.
13. Complexity studies Interdisciplinary (ecology,
social psychology, systems
analysis)
Study of how individuals, groups, and
organizations emerge, evolve, and adapt to
their environment.
Creativity, emergence, and adaptation

Innovation in Service Organizations 589
1. Rural sociology, for which Everett Rogers (1995) first developed the
concept of diffusion of innovations: In this concept, innovations
were defined as ideas or practices perceived as new by practi-
tioners (in this case, farmers). Diffusion was seen as the spread
of ideas among individuals, largely by imitation. Interventions
aimed at spreading innovation harnessed the interpersonal influ-
ence of opinion leaders and change agents, and research mapped
the social networks and adoption decisions of targeted individuals.
2. Medical sociology,inwhich similar concepts and theoretical expla-
nations were applied to doctors’ clinical behavior (most notably,
the 1966 study by Coleman, Katz, and Menzel on the spread of
prescribing of newly introduced antibiotics): Early studies in med-
ical sociology set the foundations for network analysis—the sys-
tematic study of “who knows whom” and “who copies whom”—
and led to the finding that well-networked individuals are gen-
erally better educated, have a higher social status, and are earlier
adopters of innovations (Burt 1973).
3. Communication studies,inwhich innovations were conceptualized as
new information (often “news”), and spread was seen as the trans-
mission of this information by either mass media or interpersonal
communication: Research measured the speed and direction of
the message’s transmission and studied the impact of altering key
variables such as the style of message, the communication chan-
nel (spoken, written, etc.), and the nature of exposure (Rogers and
Kincaid 1981).
4. Marketing,inwhich innovations were conceptualized as products
or services, and the adoption decision was seen as a rational (quasi-
economic) analysis of costs and benefits: Research measured the
success of efforts to increase the perceived benefits or reduce the
perceived costs of an innovation in the eyes of potential adopters.
An important stream of research in this area centered on devel-
oping mathematical models to predict adoption behavior (Bass
1969).
These early studies produced some robust empirical findings (dis-
cussed later) on the attributes of innovations, the characteristics and
behavior of adopters, and the nature and extent of interpersonal and
mass media influence on adoption decisions. But the work had a num-
ber of theoretical limitations, notably the erroneous assumptions that

590 Trisha Greenhalgh et al.
(1) the only relevant unit of analysis is the individual innovation and/or
the individual adopter; (2) an innovation is necessarily better than what
has gone before and adoption is more worthy of study than is nonadoption
or rejection; (3) patterns of adoption reflect fixed personality traits; and
(4) the findings of diffusion research are invariably transferable to new
contexts and settings. Research areas that emerged as developments—
and sometimes as breakaways—from such conceptual models
include
•Development studies,inwhich research on the spread of innovations
was explicitly broadened to include an exploration of the polit-
ical, technological, and ideological context of the innovation and
any dissemination program, and of particular innovations’ different
meaning and social value in different societies: Dif fusion of innova-
tions was reframed as centrally pertaining to the appropriateness of
particular technologies and ideas for particular situations at partic-
ular stages in development. Two important contributions from this
tradition have been (1) that the meaning of an innovation for the
agency that introduces it may be very different from that held by
the intended adopters and (2) that “innovation-system fit” (related
to the interaction between the innovation and its potential context)
is generally a more valid and useful construct than “innovation at-
tributes” (often assumed to be fixed properties of the innovation in
any context) (Bourdenave 1976).
•Health promotion,inwhich innovations were def ined as good ideas for
healthy behaviors and lifestyles, and the spread of such innovations
was expressed as the reach and uptake of health promotion programs
in defined target groups: Health promotion research has tradition-
ally used social marketing, developed from marketing theory, as
its theoretical basis. More recently, a more radical “developmental”
agenda has emerged in health promotion, with parallels to devel-
opment studies, in which a one-way transmission of advice from
the change agency to the target group has been replaced with var-
ious models of partnership and community development (Potvin,
Haddad, and Frohlich 2001).
•Evidence-based medicine,inwhich innovations were defined as health
technologies and practices supported by sound research evidence:
Until recently, the spread of innovation in this tradition was seen
as a linear and technical process at the level of the individual and

Innovation in Service Organizations 591
hence was described as changes in clinicians’ behavior in line with
evidence-based guidelines (Granados et al. 1997). Many evidence-
based medicine researchers subsequently (and perhaps somewhat be-
latedly) recognized that the implementation of most clinical guide-
lines requires changing the system and, hence, organizational as
well as individual change (Grimshaw et al. 2004). A more recent
conceptual development is the notion that the evidence base for par-
ticular technologies and practices is often ambiguous and contested
and must be continually interpreted and reframed in accordance
with the local context and priorities, a process that often involves
power struggles among various professional groups (Ferlie et al.
2001).
In the organization and management literature, we found the follow-
ing areas that were relevant to our review:
•• Studies of the structural determinants of organizational innovativeness,in
which innovation was seen as a product or process likely to make an
organization more profitable: Organizational innovativeness was
regarded as primarily influenced by structural determinants, es-
pecially size, functional differentiation (an internal division of la-
bor), slack resources, and specialization (the organization has a clear
“niche” in which it offers expertise and specialist resources). In this
area, research focuses on collecting quantitative data about the for-
mal structures of organizations, usually by sending questionnaires
to the chief executive. Such studies were among the few in our re-
view that were amenable to meta-analysis (Damanpour 1991, 1992,
1996).
•Studies of organizational process, context, and culture, whose research
focus was the adoption, assimilation, and routinization of an in-
novation: Here, the exploration of an organization’s innovativeness
concentrated on the “softer,” nonstructural aspects of its makeup,
especially the prevailing culture and climate, notably in relation
to leadership style, power balances, social relations, and attitudes
toward risk taking. This area used mainly qualitative (often ethno-
graphic) methods and centered on people and their relationships
and behavior. This research often overlapped with the mainstream
change management literature, in addition to a distinct innovation
subarea (Kanter 1988; Van de Ven et al. 1999).

592 Trisha Greenhalgh et al.
•Interorganizational studies, which examine an organization’s innova-
tiveness in relation to the influence of other organizations, partic-
ularly interorganizational communication, collaboration, competi-
tion, and norm setting: This area applied social network theory (the
notion that people are “networked” to friends and colleagues and
that these networks form channels of communication and influence
[Granovetter and Soong 1983]) to the level of the organization (e.g.,
the concept of the opinion-leading organization was introduced
and explored). Interorganizational norms (“fads and fashions”) were
seen as a key mechanism for spreading ideas among organizations
(Abrahamson 1991; Abrahamson and Fairchild 1999).
•Knowledge-based approaches to innovation in organizations,inwhich both
innovation and diffusion were radically redefined as the construc-
tion and distribution of knowledge (Nonaka and Takeuchi 1995):
A critical new concept was the organization’s absorptive capacity
for new knowledge. Absorptive capacity is a complex construct incor-
porating the organization’s existing knowledge base, “learning or-
ganization” values and goals (i.e., those that are explicitly directed
to capturing, sharing, and creating new knowledge), technolog-
ical infrastructure, leadership and knowledge sharing, and effec-
tive boundary-spanning roles with other organizations (Zahra and
George 2002).
•Narrative organizational studies,inwhich one important dimen-
sion of organizational innovativeness—the generation of ideas—was
viewed as the creative imagination of individuals in the organiza-
tion: In this field, an innovative organization is one in which new
stories can be told and that has the capacity to capture and circu-
late these stories (Czarniawska 1998; Gabriel 2000). This research
area emphasizes the rule-bound, inherently conservative nature of
large professional bureaucracies and celebrates stories for their in-
herent subversiveness. Because the principal constructions in stories
are surprise, tension, dissent, and “twists in the plot,” and because
characters can be assigned positive virtues such as honesty, courage,
or determination, stories can offer “permission to break the rules”
(Buckler and Zein 1996). In the narrative tradition, the diffusion
of innovations within organizations gives a shared story a new end-
ing. Hence, interventions to support innovation are directed toward
supporting “communities of practice” with a positive story to tell
(Bate 2004).

Innovation in Service Organizations 593
Natural,
emergent
Social Technical Managerial
Assumed Mechanism
Unpredictable,
unprogrammed,
uncertain, emergent,
adaptive, self-
organizing
Negotiated,
influenced,
enabled
Scientific, orderly,
planned, regulated,
programmed,
systems ‘‘properly
managed’’
Defining Features
Emergence,
adaptation
Dissemination,
cascading
Re-
engineering
Knowledge
transfer
NegotiationDiffusionKnowledge
construction,
making sense
Metaphor for Spread
“Let it
happen”
“Help it
happen”
“Make it
happen”
figure 2. Different Conceptual and Theoretical Bases for the Spread of In-
novation in Service Organizations
•Complexity studies are derived from general systems theory and re-
gard innovation as the emergent continuity and transformation of
patterns of interaction, understood as complex responses of humans
relating to one another in local situations: The diffusion of innova-
tions is seen as a highly organic and adaptive process in which the
organization adapts to the innovation and the innovation is adapted
to the organization (Fonseca 2001). As Figure 2 shows, this organic,
adaptive process is not easily—and perhaps not at all—controlled
by external change agencies (Plsek 2003).
One other relevant area in the organization and management lit-
erature is organizational psychology, in which innovativeness is seen
as dependent on good leadership, sound decision making, and effec-
tive human resource management (especially the motivation, training,
and support of staff ). We did not explore this literature in detail, as
it was the subject of several other projects funded by the UK De-
partment of Health Service Delivery and Organization Programme (see
www.sdo.lshtm.ac.uk/changemanagement.htm).

594 Trisha Greenhalgh et al.
A Model of Diffusion in Service
Organizations
Figure 3 shows the unifying conceptual model that we derived from
our synthesis of theoretical and empirical findings. As noted later, the
model is intended mainly as a memory aide for considering the different
aspects of a complex situation and their many interactions. It should
not be viewed as a prescriptive formula. The next section presents the
principal empirical findings from across the dif ferent research traditions,
organized broadly around the model’s main components.
The Innovation
Individual people adopt different innovations and then spread them at
different rates to other individuals. Some innovations are never adopted
at all; others are subsequently abandoned. A very extensive evidence
base from sociology (including medical sociology) supports the notion
of key attributes of innovations (as perceived by prospective adopters),
which explain much of the variance in innovations’ adoption rates. In
addition to Rogers’s authoritative review (1995), the following conclu-
sions are based on a number of more recent empirical studies of ser-
vice innovations in health care (see Greenhalgh et al. 2005a for the full
references):
Relative Advantage. Innovations that have a clear, unambiguous ad-
vantage in either effectiveness or cost-effectiveness are more easily
adopted and implemented (for strong indirect and moderate direct
evidence, see Dirksen, Ament, and Go 1996; Marshall 1990; Meyer,
Johnson, and Ethington 1997; and Rogers 1995). If potential users see
no relative advantage in the innovation, they generally will not consider
it further; in other words, relative advantage is a sine qua non for adop-
tion (for strong direct and moderate indirect evidence, see Rogers 1995).
Nevertheless, relative advantage alone does not guarantee widespread
adoption (for strong direct evidence, see Denis et al. 2002; Fitzgerald
et al. 2002; and Grimshaw et al. 2004). Even so-called evidence-based
innovations undergo a lengthy period of negotiation among potential
adopters, in which their meaning is discussed, contested, and reframed.
Such discourse can increase or decrease the innovation’s perceived relative
advantage (for moderate direct evidence, see Ferlie et al. 2001).

Innovation in Service Organizations 595
COMMUNICATION
AND INFLUENCE
DIFFUSION
(informal, unplanned)
Social networks
Homophily
Peer opinion
Marketing
Expert opinion
Champions
Boundary spanners
Change agents
DISSEMINATION
(formal, planned)
SYSTEM READINESS
FOR INNOVATION
Tension for change
Innovation-system fit
Power balances
(supporters v. opponents)
Assessment of implications
Dedicated time/resour ces
Monitoring and feedback
OUTER CONTEXT
Sociopolitical climate
Incentives and mandates
Interorganizational
norm-setting and
networks
Environmental stability
User system
Adoption/assimilation
System antecedents
System readiness
Implementation
Consequences
Outer context
Knowledge
purveyors
Resource system
The innovation
Change agency
Dissemination
Diffusion
LINKAGE
LINKAGE
LINKAGE
Design stage Implementation stage
Shared meanings and mission Communication and information
Effective knowledge transfer User orientation
User involvement in specification Product augmentation, e.g. technical help
Capture of user-led innovation Project management support
IMPLEMENTATION
PROCESS
Decision maki ng devolved
to frontline teams
Hands-on approach by
leaders and managers
Human resource issues,
especially trai ning
Dedicated resources
Internal communication
External collaboration
Reinvention/developmen t
Feedback on progress
SYSTEM ANTECEDENTS FOR INNOVATION
Structure
Size/maturity
Formalization
Differentiation
Decentralization
Slack resources
Receptive context for change
Leadership and vision
Good managerial relations
Risk-taking climate
Clear goals and priorities
High-quality data capture
Absorptive capacity for new knowledg e
Preexisting knowledge/skills base
Ability to find, interpret, recodify,
and integrate new knowledge
Enablement of knowledge sharing
via internal and external networks
ADOPTER
Needs
Motivation
Values and goals
Skills
Learning style
Social networks
THE INNOVATION
Relative advantage
Compatibility
Low complexity
Trialability
Observability
Potential for reinvention
Fuzzy boundaries
Risk
Task issues
Nature of knowledge
required (tacit/explicit)
Technical support
ASSIMILATION
Complex, nonlinear
process
“Soft periphery” elements
figure 3. Conceptual Model for Considering the Determinants of Diffusion, Dissemination, and Implementation of Innovations
in Health Service Delivery and Organization, Based on a Systematic Review of Empirical Research Studies

596 Trisha Greenhalgh et al.
Compatibility. Innovations that are compatible with the intended
adopters’ values, norms, and perceived needs are more readily adopted (for
strong direct evidence, see Aubert and Hamel 2001; Denis et al. 2002;
Ferlie et al. 2001; Foy et al. 2002; and Rogers 1995). Compatibility
with organizational or professional norms, values, and ways of working
is an additional determinant of successful assimilation (for strong direct
evidence, see Denis et al. 2002; Fennell and Warnecke 1988; and Ferlie
et al. 2001).
Complexity. Innovations that are perceived by key players as simple
to use are more easily adopted (for strong direct evidence, see Denis
et al. 2002; Grilli and Lomas 1994; Marshall 1990; Meyer and Goes
1988; Meyer, Johnson, and Ethington 1997; and Rogers 1995). Per-
ceived complexity can be reduced by practical experience and demon-
stration (for moderate direct evidence, see Plsek 2003). If the innovation
can be broken down into more manageable parts and adopted incremen-
tally, it will be more easily adopted (for strong indirect and moderate
direct evidence, see Plsek 2003; and Rogers 1995). If an innovation in
an organizational setting has few response barriers that must be over-
come, it will be assimilated more easily (for strong indirect and moderate
direct evidence, see Rogers 1995). Interventions to reduce the number
and extent of such response barriers improve the chances of successful
adoption (limited evidence).
Trialability. Innovations with which the intended users can exper-
iment on a limited basis are adopted and assimilated more easily (for
strong direct evidence, see Grilli and Lomas 1994; Plsek 2003; Rogers
1995; and Yetton, Sharma, and Southon 1999). Such experimentation
can be encouraged by providing “trialability space” (for strong indirect
and moderate direct evidence, see Øvretveit et al. 2002; Plsek 2003; and
Rogers 1995).
Observability. If the benefits of an innovation are visible to intended
adopters, it will be adopted more easily (for strong direct evidence, see
Denis et al. 2002; Grilli and Lomas 1994; Meyer and Goes 1988; and
Øvretveit et al. 2002). Initiatives to make more visible the benefits of
an innovation (e.g., through demonstrations) increase the likelihood of
their assimilation (limited evidence).
Reinvention. If potential adopters can adapt, refine, or otherwise mod-
ify the innovation to suit their own needs, it will be adopted more
easily (for strong direct evidence, see Meyer, Johnson, and Ethington
1997; and Rogers 1995). Reinvention is especially important to those

Innovation in Service Organizations 597
innovations that arise spontaneously as “good ideas in practice” and
spread through informal, decentralized, horizontal social networks (for
moderate indirect evidence, see Rogers 1995; see also “fuzzy boundaries”
in this article).
These “standard” attributes (which, apart from reinvention, are ex-
tensively cited) are necessary but not sufficient to explain the adoption
and assimilation of complex innovations in organizations. Additional
key attributes are as follows (note that for clarity we have conflated some
that researchers considered separately):
Fuzzy Boundaries. Complex innovations in service organizations can
be conceptualized as having a “hard core” (the irreducible elements of the
innovation itself ) and a “soft periphery” (the organizational structures
and systems required for the full implementation of the innovation); the
adaptiveness of the “soft periphery” is a key attribute of the innovation
(for moderate direct evidence, see Denis et al. 2002). The concept of a soft
periphery links with Rogers’s aforementioned concept of reinvention and
with “innovation-system fit” as an important feature of system readiness.
Risk. If the innovation carries a high degree of uncertainty of out-
come that the individual perceives as personally risky, it is less likely to
be adopted (for strong direct evidence, see Meyer and Goes 1988; and
Meyer, Johnson, and Ethington 1997). Because an innovation’s risks and
benefits are not evenly distributed in an organization, the more the bal-
ance between risks and benefits reflects the organization’s power base,
the more likely the innovation is to be assimilated (for moderate direct
evidence, see Denis et al. 2002; and Ferlie et al. 2001).
Task Issues. If the innovation is relevant to the performance of the
intended user’s work and if it improves task performance, it will be
adopted more easily (for moderate direct evidence, see Yetton, Sharma,
and Southon 1999). Interventions to enhance task relevance improve the
chances of successful adoption (limited evidence). If the innovation is
feasible, workable, and easy to use, it will be adopted more easily (for
strong direct evidence, see Dobbins, Cockerill, and Barnsley 2001; Foy
et al. 2002; Meyer and Goes 1988; and Yetton, Sharma, and Southon
1999). Interventions to improve the feasibility and workability of inno-
vations for key staff members and teams improve the chances of successful
adoption (limited evidence).
Knowledge Required to Use It. If the knowledge required for the inno-
vation’s use can be codified and transferred from one context to another,
it will be adopted more easily (for strong indirect and moderate direct

598 Trisha Greenhalgh et al.
evidence, see Adler, Kwon, and Singer 2003; Aubert and Hamel 2001;
and O’Neill, Pouder, and Buchholtz 2002).
Augmentation/Support. If a technology is supplied as an “augmented
product” (e.g., with customization, training, and a help desk), it will be
assimilated more easily (for strong moderate direct evidence, see Aubert
and Hamel 2001).
Our full report gives a number of examples of studies that failed
to support the importance of even the most well-established attributes
in certain settings (Greenhalgh et al. 2005a). This finding illustrates
the important principle that the attributes are neither stable features
of the innovation nor sure determinants of their adoption or assimila-
tion. Rather, it is the interaction among the innovation, the intended
adopter(s), and a particular context that determines the adoption rate.
As Dearing and And commented:
Conceptualising innovations as “having” attributes is a common
heuristic that people employ when they are judging something new.
Yet this tendency serves to obscure the importance of human percep-
tion in the diffusion of innovations. What is new to one person may
be “old” to another.... Moreover, the decision to adopt and/or use
the innovation is based on individual perceptions of the innovation’s
worth relative to other ways of accomplishing the same goal. What
is easy for one person to use may be exceedingly difficult for another.
(Dearing and And 1994, 19)
Adoption by Individuals
People are not passive recipients of innovations. Rather (and to a greater
or lesser extent in different persons), they seek innovations, experiment
with them, evaluate them, find (or fail to find) meaning in them, de-
velop feelings (positive or negative) about them, challenge them, worry
about them, complain about them, “work around” them, gain experience
with them, modify them to fit particular tasks, and try to improve or re-
design them—often through dialogue with other users. This diverse list
of actions and feelings highlights the complex nature of adoption as a pro-
cess and contrasts markedly with the widely cited “adopter categories”
(“early adopter,” “laggard”) that have been extensively misapplied as
explanatory variables. There is little empirical support for these stereo-
typical and value-laden terms, which fail to acknowledge the adopter
as an actor who interacts purposefully and creatively with a complex
innovation.

Innovation in Service Organizations 599
The seven aspects of adopters and the adoption process that we used in
our overall model are based on Rogers’s extensive overview of the wider
literature on adoption (1995) plus additional empirical studies of health
service innovations (see full report for details).
General Psychological Antecedents. We identified a large literature from
cognitive and social psychology on individual traits associated with the
propensity to try out and use innovations (e.g., tolerance of ambiguity,
intellectual ability, motivation, values, and learning style). This evidence
has been largely ignored by researchers studying the diffusion of innova-
tions and was beyond the scope of our own study, but it is ripe for review
in relation to this research question.
Context-Specific Psychological Antecedents. An intended adopter who is
motivated and able (in terms of values, goals, specific skills, and so on)
to use a particular innovation is more likely to adopt it (for strong direct
evidence, see Ferlie et al. 2001; Gladwin, Dixon, and Wilson 2002;
and Yetton, Sharma, and Southon 1999). If the innovation meets an
identified need by the intended adopter, he or she is more likely to adopt
it (for strong indirect evidence, see Hall and Hord 1987; and Wejnert
2002).
Meaning. The meaning of the innovation for the intended adopter
has a powerful influence on the adoption decision (for strong indirect
and moderate direct evidence, see Dearing and And 1994; and Timmons
2001). If the meaning attached to the innovation by individual adopters
matches the meaning attached by top management, service users, and
other stakeholders, the innovation is more likely to be assimilated (for
moderate indirect evidence, see Eveland 1986). The meaning attached to
an innovation is generally not fixed but can be negotiated and reframed,
for example, through discourse within the organization or across inter-
organizational networks (for strong direct evidence, see Ferlie et al.
2001). The success of initiatives to support such a reframing of meaning
is variable and not easy to predict (limited evidence).
The Adoption Decision. The decision by an individual within an orga-
nization to adopt a particular innovation is rarely independent of other
decisions. It may be contingent (dependent on a decision made by some-
one else in the organization), collective (the individual has a “vote” but
ultimately must acquiesce to the decision of a group), or authoritative
(the individual is told whether or not to adopt it) (Rogers 1995). Au-
thoritative decisions (e.g., making adoption by individuals compulsory)
may increase the chance of initial adoption by individuals but may also

600 Trisha Greenhalgh et al.
reduce the chance that the innovation is successfully implemented and
routinized (for moderate indirect evidence, see Rogers 1995).
Adoption is a process rather than an event, with different concerns
being dominant at different stages. The adoption process in individ-
uals is traditionally presented as having five stages: awareness, per-
suasion, decision, implementation, and confirmation (Rogers 1995).
However, we found that a lesser-known model, the Concerns Based
Adoption Model developed for innovation in schools, better explained
the findings of empirical studies of complex service innovations in an
organizational context. This model provided three components for our
model:
Concerns in Preadoption Stage. Important prerequisites for adoption are
that the intended adopters are aware of the innovation; have sufficient
information about what it does and how to use it; and are clear about
how the innovation would affect them personally, for example, in terms
of costs (for strong indirect evidence, see Hall and Hord 1987).
Concerns during Early Use. Successful adoption is more likely if the
intended adopters have continuing access to information about what the
innovation does and to suff icient training and support on task issues (i.e.,
about fitting the innovation to daily work) (for strong indirect evidence,
see Hall and Hord 1987).
Concerns in Established Users. Successful adoption is more likely if
adequate feedback is provided to the intended adopters about the con-
sequences of adoption (for strong indirect evidence, see Hall and Hord
1987) and if the intended adopters have sufficient opportunity, auton-
omy, and support to adapt and ref ine the innovation to improve its fitness
for purpose (for strong indirect evidence, see Rogers 1995).
Assimilation by the System
Most of the research on the diffusion of innovations focused on simple,
product-based innovations, for which the unit of adoption is the indi-
vidual, and diffusion occurs by means of simple imitation (Rogers 1995).
It is important not to use this literature to overgeneralize to complex,
process-based innovations in service organizations, for which the unit
of adoption (at this level, more often called assimilation)isthe team,
department, or organization in which various changes in structures or
ways of working will be required. In such circumstances, there is almost
always a formal decision-making process, an evaluation phase or phases,

Innovation in Service Organizations 601
and planned and sustained efforts at implementation. In other words,
empirical work in the field of organization and management clearly
shows that successful individual adoption is only one component of the
assimilation of complex innovations in organizations. The evaluation of
system readiness and the crucial implementation phase are considered
separately later, but one overarching concept should be borne in mind
about the assimilation process as a whole:
The Assimilation. Although one large, high-quality study (Meyer and
Goes 1988) demonstrated an organizational parallel to the “stages” of
individual adoption, comprising “knowledge-awareness,” “evaluation-
choice,” and “adoption-implementation,” the remaining empirical evi-
dence was more consistent with an organic and often rather messy model
of assimilation in which the organization moved back and forth between
initiation, development, and implementation, variously punctuated by
shocks, setbacks, and surprises (for strong direct evidence, see Van de
Venetal. 1999).
Diffusion and Dissemination
The various influences that help spread the innovation can be thought of
as lying on a continuum between pure diffusion (in which the spread of
innovations is unplanned, informal, decentralized, and largely horizon-
tal or mediated by peers) and active dissemination (in which the spread
of innovation is planned, formal, often centralized, and likely to occur
more through vertical hierarchies; see Figure 2). Whereas mass media
and other impersonal channels may create awareness of an innovation,
interpersonal influence through social networks (defined as “the pattern
of friendship, advice, communication and support which exists among
members of a social system” [Valente 1996, 70]) is the dominant mecha-
nism for diffusion. Again drawing on Rogers’s overview (1995) as well as
other empirical work (see full report for more references), we identified
a number of components for our model:
Network Structure. The adoption of innovations by individuals is
powerfully influenced by the structure and quality of their social net-
works (for strong indirect and moderate direct evidence, see Fennell and
Warnecke 1988; Valente 1996; and West et al. 1999). Different groups
have different types of social networks. Doctors, for example, tend to
operate in informal, horizontal networks, and nurses more often have
formal, vertical networks (for moderate direct evidence, see West et al.

602 Trisha Greenhalgh et al.
1999). Different social networks also have different uses for different
types of influence; for example, horizontal networks are more effective
for spreading peer influence and supporting the construction and re-
framing of meaning; vertical networks are more effective for cascading
codified information and passing on authoritative decisions (for moderate
indirect evidence and limited direct evidence, see Rogers 1995; and West
et al. 1999).
Homophily. The adoption of innovations by individuals is more likely
if they are homophilous—that is, have similar socioeconomic, educational,
professional, and cultural backgrounds—with current users of the in-
novation (for strong direct evidence, see Fennell and Warnecke 1988;
Fitzgerald et al. 2002; and West et al. 1999).
Opinion Leaders. Some persons have a particular influence on the be-
liefs and actions of their colleagues (for strong direct evidence, see Becker
1970; and Coleman, Katz, and Menzel 1966). Expert opinion leaders ex-
ert influence through their authority and status, and peer opinion leaders
exert influence through their representativeness and credibility (for mod-
erate direct evidence, see Fitzgerald et al. 2002; and Locock et al. 2001).
Opinion leaders can have either a positive or negative influence. If a
project is insufficiently appealing (e.g., in clarity of goals, organization,
and resources), it will not attract the support of key opinion leaders (for
strong indirect and moderate direct evidence, see Locock et al. 2001;
Rogers 1995).
Harnessing the Opinion Leader’s Influence. Even though the powerful
impact of social influence (such as that of opinion leaders) in natural-
istic settings is well established, attempts to engage such individuals
in planned change efforts have often had disappointing results. In cases
in which opinion leaders have been trained to influence the behavior of
their peers (e.g., to persuade fellow clinicians to follow a new guideline),
the impact is generally positive in direction but small in magnitude (for
strong direct evidence, see Thomson O’Brien et al. 2003). The failure to
identify the true opinion leaders and, in particular, the failure to distin-
guish between monomorphic opinion leaders (influential for a particular
innovation only) and polymorphic opinion leaders (influential across a
wide range of innovations) may limit the success of such intervention
strategies (for strong indirect and moderate direct evidence, see Locock
et al. 2001; and Rogers 1995).
Champions. The adoption of an innovation by individuals in an or-
ganization is more likely if key individuals in their social networks

Innovation in Service Organizations 603
are willing to support the innovation (for strong indirect and moderate
direct evidence, see Backer and Rogers 1998; Markham 1998; Meyer and
Goes 1988; and Schon 1963). The different champion roles for organiza-
tional innovations include (1) the organizational maverick, who gives the
innovators autonomy from the organization’s rules, procedures, and sys-
tems so they can establish creative solutions to existing problems; (2) the
transformational leader, who harnesses support from other members of
the organization; (3) the organizational buffer, who creates a loose mon-
itoring system to ensure that innovators properly use the organization’s
resources while still allowing them to act creatively; and (4) the network
facilitator, who develops cross-functional coalitions within the organiza-
tion (for moderate indirect evidence, see Shane 1995). There is very little
direct empirical evidence on how to identify, and systematically harness
the energy of, organizational champions.
Boundary Spanners. An organization is more likely to adopt an in-
novation if those people who have significant social ties both inside
and outside the organization are able and willing to link the organiza-
tion to the outside world in relation to this particular innovation. Such
individuals play a pivotal role in capturing the ideas that will become or-
ganizational innovations (for strong indirect evidence, see Rogers 1995;
for moderate direct evidence, see Kimberly and Evanisko 1981). Or-
ganizations that promote and support the development and execution
of boundary-spanning roles are more likely to become aware of and as-
similate innovations quickly (for moderate direct evidence, see Barnsley,
Lemieux-Charles, and McKinney 1998; Ferlie et al. 2001; and Tushman
1977).
Formal Dissemination Programs. When a planned dissemination pro-
gram is used for the innovation (e.g., led by an external change agency),
it will be more effective if the program’s organizers (1) take full account
of potential adopters’ needs and perspectives, with particular attention to
the balance of costs and benefits for them; (2) tailor dif ferent strategies to
the different demographic, structural, and cultural features of different
subgroups; (3) use a message with appropriate style, imagery, metaphors,
and so on; (4) identify and use appropriate communication channels; and
(5) incorporate rigorous evaluation and monitoring of defined goals and
milestones (for strong indirect evidence, see Rogers 1995).
The diverse literature on diffusion and dissemination highlighted
an important area of contestation in paradigms of diffusion. Most dif-
fusion research has addressed proactively developed innovations (e.g.,

604 Trisha Greenhalgh et al.
technologies or products developed in formal research programs) whose
main mechanism of spread is centrally driven and controlled (what we
have defined as dissemination). But many innovations in service delivery
and organization occur as “good ideas” in local services, which spread
informally and in a largely uncontrolled way (diffusion). This tension,
which has received remarkably little attention in the literature we re-
viewed, is illustrated in Figure 2.
System Antecedents for Innovation
Different organizations provide widely differing contexts for innova-
tions, and some features of organizations (both structural and “cultural”)
have been shown to influence the likelihood that an innovation will
be successfully assimilated (i.e., adopted by all relevant individuals and
incorporated into “business as usual”).
Structural Determinants of Innovativeness. We identified four previ-
ous meta-analyses that included both manufacturing and service orga-
nizations (Damanpour 1991 [see Table 2], 1992, 1996; Tornatsky and
Klein 1982) and 15 additional empirical studies (examined in 17 pa-
pers) from the service sector (Anderson and West 1998; Baldridge and
Burnham 1975; Burns and Wholey 1993; Castle 2001; Champagne
et al. 1991; Dopson et al. 2002; Dufault et al. 1995; Fitzgerald et al.
2002; Goes and Park 1997; Gosling, Westbrook, and Braithwaite 2003;
Kimberly and Evanisko 1981; Meyer and Goes 1988; Newton et al.
2003; Nystrom, Ramamurthy, and Wilson 2002; Patel 1996; Rashman
and Hartley 2002; Wilson, Ramamurthy, and Nystrom 1999). Their
findings vary somewhat, though less so than is often claimed. They
suggest that an organization will assimilate innovations more readily if
it is large, mature, functionally differentiated (i.e., divided into semi-
autonomous departments and units), and specialized, with foci of profes-
sional knowledge; if it has slack resources to channel into new projects;
and if it has decentralized decision-making structures (strong direct evi-
dence). Size is almost certainly a proxy for other determinants, including
slack resources and functional differentiation.
Although these structural determinants are significantly, positively,
and consistently associated with organizational innovativeness, together
they account for less than 15 percent of the variation among compa-
rable organizations. Furthermore, the relationship between structural
determinants and innovativeness is moderated by and/or contingent on

Innovation in Service Organizations 605
TABLE 2
Impact of Structural Determinants on Organizational Innovativeness from a
Meta-Analysis of 23 Studies of Service and Manufacturing
Association with
Potential Organizational
Determinants Definition Innovativeness
Administrative
intensity Indicator of administrative overhead. Positive, significant
Centralization Extent to which decision-making
autonomy is dispersed or
concentrated in an organization.
Negative, significant
Complexity “Specialization,” “functional
differentiation,” and
“professionalism.”
Positive, significant
External
communication Degree of organization members’
involvement and participation in
extraorganizational professional
activities.
Positive, significant
Formalization Reflects emphasis on following rules
and procedures in conducting
organizational activities.
No significant
association
Functional
differentiation Extent to which divided into
different units. Positive, significant
Internal
communication Extent of communication among
organizational units. Positive, significant
Managerial
attitude toward
change
Extent to which managers or
members of the dominant coalition
favor change.
Positive, significant
Managerial tenure Length of managers’ service and
experience within an organization. No significant
association
Professionalism Professional knowledge of an
organization’s members. Positive, significant
Slack resources Reflects an organization’s resources
beyond minimal requirement to
maintain operations.
Positive, significant
Specialization Number of an organization’s
specialties. Positive, significant
Technical capacity Reflects an organization’s technical
resources and technical potential. Positive, significant
Vertical
differentiation Number of levels in an organization’s
hierarchy. No signif icant
association
Source: Damanpour 1991.
additional factors (e.g., the radicalness of the innovation, whether it is
administrative or technical, and the stage of adoption). There is little
empirical evidence to support the efficacy of interventions to change
an organization’s structure to make it more “innovative,” except that

606 Trisha Greenhalgh et al.
establishing semiautonomous multidisciplinary project teams is inde-
pendently associated with successful implementation.
One important weakness of the literature on structural determinants
of innovativeness is the assumption that they can be treated as variables
whose impact can be isolated and independently quantified. For exam-
ple, the empirical studies of organizational size implicitly assume that
there is a “size effect” that is worth measuring and that is to some extent
generalizable. An alternative theoretical approach (House, Rousseau, and
Thomas-Hunt 1995), supported by a number of recent detailed quali-
tative studies (Champagne et al. 1991; Ferlie et al. 2001), is that the
determinants of organizational innovativeness interact in a complex, un-
predictable, and nongeneralizable way with one another.
There is consistent empirical evidence for two more nonstructural
determinants of organizational innovativeness:
Absorptive Capacity for New Knowledge. An organization that is system-
atically able to identify, capture, interpret, share, reframe, and recodify
new knowledge; to link it with its own existing knowledge base; and to
put it to appropriate use will be better able to assimilate innovations,
especially those that include technologies (for strong direct evidence,
see Barnsley, Lemieux-Charles, and McKinney 1998; and Ferlie et al.
2001). Prerequisites for absorptive capacity include the organization’s
existing knowledge and skills base (especially its store of tacit, uncodi-
fiable knowledge) and preexisting related technologies, a “learning or-
ganization” culture, and proactive leadership directed toward sharing
knowledge (for strong direct evidence, see Barnsley, Lemieux-Charles,
and McKinney 1998; Ferlie et al. 2001; and Zahra and George 2002).
The knowledge that underpins the adoption, dissemination, and imple-
mentation of a complex innovation within an organization is not objec-
tive or given. Rather, it is socially constructed and frequently contested
and must be continually negotiated among members of the organization
or system. Strong, diverse, and organic (i.e., flexible, adaptable, and lo-
cally grown) intraorganizational networks (especially opportunities for
interprofessional teamwork, and the involvement of clinicians in man-
agement networks and vice versa) help this process and facilitate the
development of shared meanings and values in relation to the innova-
tion (for moderate direct evidence, see Barnsley, Lemieux-Charles, and
McKinney 1998; and Ferlie et al. 2001).
An important use of knowledge in health care organizations is the
application of research evidence for the efficacy of health technologies.
Health professionals should ensure that they and their staff are aware of

Innovation in Service Organizations 607
new developments (and new definitions of what is obsolete) in diagnostic
tests, drugs, surgical procedures, and so on, and modify their practice
accordingly. A major overview of high-quality qualitative studies of how
research evidence is identified, circulated, evaluated, and used in health
care organizations (Dopson et al. 2002) confirms other findings from
the mainstream knowledge-utilization literature, which suggest that
before it can contribute to organizational change initiatives, knowledge
must be enacted and made social, entering into the stock of knowledge
constructed and shared by other individuals. Knowledge depends for its
circulation on interpersonal networks and will spread only if these social
features are taken into account and barriers are overcome.
Receptive Context for Change. The receptive context for change incor-
porates a number of organizational features that have been independently
associated with its ability to embrace new ideas and face the prospect of
change (Pettigrew and McKee 1992). An organization with such a recep-
tive context will be better able to assimilate innovations. In addition to
an absorptive capacity for new knowledge, the components of receptive
context include strong leadership, clear strategic vision, good manage-
rial relations, visionary staff in pivotal positions, a climate conducive
to experimentation and risk taking, and effective data capture systems
(for strong indirect and moderate direct evidence, see Anderson and
West 1998; Barnsley, Lemieux-Charles, and McKinney 1998; Dopson
et al. 2002; Gosling, Westbrook, and Braithwaite 2003; Newton et al.
2003; Nystrom, Ramamurthy, and Wilson 2002; Pettigrew and McKee
1992; and Van de Ven et al. 1999). Leadership may be especially helpful
in encouraging organizational members to break out of the convergent
thinking and routines that are the norm in large, well-established orga-
nizations (for strong direct evidence, see Van de Ven et al. 1999).
System Readiness for Innovation
An organization may be amenable to innovation in general but not
ready or willing to assimilate a particular innovation. As Figure 3
shows, formal consideration of the innovation allows the organization
to move (or perhaps choose not to move) to a specific state of system
readiness for that innovation. The elements of system readiness are as
follows:
Tension for Change. If staf f perceive that the current situation is intol-
erable, a potential innovation is more likely to be assimilated successfully
(for moderate direct evidence, see Gustafson et al. 2003).

608 Trisha Greenhalgh et al.
Innovation-System Fit. An innovation that fits with the organization’s
existing values, norms, strategies, goals, skill mix, supporting technolo-
gies, and ways of working is more likely to be assimilated (for strong
indirect and moderate direct evidence, see Gustafson et al. 2003; Rogers
1995; and the related concept of “fuzzy boundaries” in this article).
Assessment of Implications. If the implications of the innovation (in-
cluding its subsequent effects) are fully assessed and anticipated, the
innovation is more likely to be assimilated (for strong indirect and mod-
erate direct evidence, see Gustafson et al. 2003; and Rogers 1995). Most
of the following implementation issues are amenable to advance assess-
ment and planning:
Support and Advocacy. If the supporters of the innovation outnumber
and are more strategically placed than its opponents are, it is more likely
to be assimilated (for strong indirect and moderate direct evidence, see
Champagne et al. 1991; Gustafson et al. 2003; Rogers 1995; and also
“champions,” in this article).
Dedicated Time and Resources. If the innovation starts out with a bud-
get and if the allocation of resources is both adequate and continuing, it
is more likely to be assimilated (for strong indirect and moderate direct
evidence, see Gustafson et al. 2003; and Rogers 1995).
Capacity to Evaluate the Innovation. If the organization has tight sys-
tems and appropriate skills in place to monitor and evaluate the impact
of the innovation (both anticipated and unanticipated), the innovation
is more likely to be assimilated and sustained (for strong indirect and
moderate direct evidence, see Gustafson et al. 2003; Plsek 2003; and
Rogers 1995).
The Outer Context: Interorganizational
Networks and Collaboration
An organization’s decision to adopt an innovation and its efforts to im-
plement and sustain it depend on a number of external influences:
Informal Interorganizational Networks. An important influence on an
organization’s decision to adopt is whether a threshold proportion of
comparable (homophilous) organizations have done so or plan to do so
(for strong direct evidence, see Burns and Wholey 1993; Fennell and
Warnecke 1988; Robertson and Wind 1983; and Westphal, Gulati, and
Shortell 1997). A “cosmopolitan” organization (one that is externally

Innovation in Service Organizations 609
well networked with others) is more susceptible to this influence (for
strong direct evidence, see Burns and Wholey 1993; Fennell and War-
necke 1988; Robertson and Wind 1983; and Westphal, Gulati, and
Shortell 1997). Interorganizational networks promote the adoption of
an innovation only after this is generally perceived as “the norm.” Until
that time, networks can also serve to dissuade organizations from adopt-
ing innovations that have no perceived advantages (for strong indirect
and moderate direct evidence, see Abrahamson 1991; Fitzgerald et al.
2002; and Westphal, Gulati, and Shortell 1997). Integrative organiza-
tional forms (such as the UK National Health Service, Health Main-
tenance Organizations, and professionally led networks of health care
providers), which link provider organizations through common man-
agement and governance structures and explicit shared values and goals,
can help spread innovations among member organizations (for moderate
direct evidence, see Meyer, Johnson, and Ethington 1997).
Intentional Spread Strategies. Formal networking initiatives such as
quality improvement collaboratives (Øvretveit et al. 2002) or “Beacon”
schemes (Rashman and Hartley 2002), aimed at sharing ideas and knowl-
edge construction, are sometimes but not always effective (for moderate
direct evidence, see Flamm, Berwick, and Kabcenell 1998; Horbar et al.
2001; Leape et al. 2000; O’Connor et al. 1996; Øvretveit et al. 2002;
Rashman and Hartley 2002; and Rogowski et al. 2001). Such initiatives
are often expensive, and the gains from them are difficult to measure;
evidence of their cost-effectiveness is limited. The greatest success fac-
tors of health care quality improvement collaboratives are (1) the nature
of the topic chosen for improvement; (2) the capacity and motivation
of participating teams, particularly their leadership and team dynam-
ics; (3) the motivation and receptivity to change of the organizations
they represent; (4) the quality of facilitation, particularly the provision
of opportunities to learn from others in an informal space; and (5) the
quality of support provided to teams during the implementation phase
(for moderate direct evidence, see Øvretveit et al. 2002).
Wider Environment. The evidence for the impact of environmental
variables on organizational innovativeness in the service sector is sparse
and heterogeneous, with each group of researchers exploring somewhat
different aspects of the “environment” or “changes in the environment.”
Environmental uncertainty has either a small positive impact or no im-
pact on innovativeness (for moderate direct evidence, see Kimberly and

610 Trisha Greenhalgh et al.
Evanisko 1981; and Meyer and Goes 1988), and there may be small
positive effects from interorganizational competition and the higher so-
cioeconomic status of patients/clients (limited evidence).
Political Directives. Although our review was not designed to tap cen-
trally into the literature on policymaking and its impact, some empirical
studies of innovation formally measured the effect of the policy context
on the adoption of a particular innovation. A policy “push” occurring at
the early stage of implementation of an innovation initiative can increase
its chances of success, perhaps most crucially by making available a ded-
icated funding stream (for strong direct evidence, see Exworthy, Berney,
and Powell 2003; Fitzgerald et al. 2002; Granados et al. 1997; and
Hughes et al. 2002). External mandates (political “must-dos”) increase
an organization’s predisposition (i.e., motivation), but not its capacity,
to adopt an innovation (for moderate direct evidence, see Taylor et al.
1998). Such mandates (or the fear of them) may divert activity away from
innovations as organizations second-guess what they will be required to
do next rather than focus on locally generated ideas and priorities (for
strong indirect evidence, see Meyers, Sivakumar, and Nakata 1999; for
moderate direct evidence, see Exworthy, Berney, and Powell 2003).
Implementation and Routinization
Meyers, Sivakumar, and Nakata define implementation as “the early us-
age activities that often follow the adoption decision” (1999, 295). The
evidence regarding the implementation of innovations was particularly
complex and relatively sparse, and it was difficult to disentangle it from
that regarding change management and organizational development in
general. Implementation depends on many of the factors already cov-
ered in relation to the initial adoption decision and the early stages of
assimilation. At the organizational level, the move from considering an
innovation to successfully routinizing it is generally a nonlinear process
characterized by multiple shocks, setbacks, and unanticipated events
(Van de Ven et al. 1999). The key components of system readiness for
an innovation are highly relevant to the early stages of implementation.
In addition, a number of additional elements are specifically associated
with successful routinization:
Organizational Structure. An adaptive and flexible organizational
structure, and structures and processes that support devolved decision
making in the organization (e.g., strategic decision making devolved

Innovation in Service Organizations 611
to departments, operational decision making devolved to teams on the
ground) enhance the success of implementation and the chances of rou-
tinization (for strong indirect and direct evidence, see Meyers, Sivakumar,
and Nakata 1999; and Van de Ven et al. 1999).
Leadership and Management. Top management support, advocacy of
the implementation process, and continued commitment to it enhance
the success of implementation and routinization (for strong indirect and
moderate direct evidence, see Green 1998; Gustafson et al. 2003; Meyers,
Sivakumar, and Nakata 1999). If the innovation aligns with the earlier
goals of both top management and middle management and if the leaders
are actively involved and frequently consulted, the innovation is more
likely to be routinized (for moderate direct evidence, see Gustafson et al.
2003).
Human Resource Issues. Successful routinization of an innovation in
an organization depends on the motivation, capacity, and competence
of individual practitioners (for moderate direct evidence, see Gustafson
et al. 2003). The early and widespread involvement of staff at all lev-
els, perhaps through formal facilitation initiatives, enhances the success
of implementation and routinization (for strong indirect evidence, see
Meyers, Sivakumar, and Nakata 1999; for moderate direct evidence, see
Kitson, Harney, and McCormack 1998). When job changes are few and
clear, high-quality training materials are available, and timely on-the-
job training is provided, successful and sustained implementation is
more likely (for strong indirect and moderate direct evidence, see Green
1998; Gustafson et al. 2003; Meyers, Sivakumar, and Nakata 1999;
and McCormick, Steckler, and Mcleroy 1995). Team-based training may
be more effective than individual training when the learning involves
implementing a complex technology (for moderate direct evidence, see
Edmondson, Bohmer, and Pisano 2001).
Funding. If there is dedicated and ongoing funding for its implemen-
tation, the innovation is more likely to be implemented and routinized
(for strong direct evidence, see Elliott et al. 1998; Fitzgerald et al. 2002;
Green 1998; Gustafson et al. 2003; and Hughes et al. 2002).
Intraorganizational Communication. Effective communication across
structural (e.g., departmental) boundaries within the organization en-
hances the success of implementation and the chances of routiniza-
tion (for strong indirect evidence, see Meyers, Sivakumar, and Nakata
1999). A narrative approach (i.e., the purposeful construction of a shared
and emergent organizational story of “what we are doing with this

612 Trisha Greenhalgh et al.
innovation”) can serve as a powerful cue to action (for moderate indirect
evidence, see Gabriel 2000; for limited direct evidence, see Bate 2004).
Interorganizational Networks. The more complex the implementation
that is needed for a particular innovation, the greater the significance
of the interorganizational network will be to the implementation’s suc-
cess (for moderate indirect evidence, see Meyers, Sivakumar, and Nakata
1999; and Valente 1995).
Feedback. Accurate and timely information about the impact of the
implementation process (through efficient data collection and review
systems) increases the chance of successful routinization (for strong in-
direct and moderate direct evidence, see Green 1998; and Grimshaw
et al. 2004).
Adaptation/Reinvention. If an innovation is adapted to the local con-
text, it is more likely to be successfully implemented and routinized (for
strong indirect and moderate direct evidence, see Gustafson et al. 2003;
Øvretveit et al. 2002; and Rogers 1995).
Linkage among Components of the Model
There is some empirical evidence (and also robust theoretical arguments)
for building strong links among the different components of the model
in Figure 3:
Linkage at the Development Stage. An innovation that is centrally de-
veloped (e.g., in a research center) is more likely to be widely and success-
fully adopted if the developers or their agents are linked with potential
users at the development stage in order to capture and incorporate the
users’ perspective (for strong indirect evidence, see Rogers 1995). Such
linkage should aim not merely for “specification” but also for a shared
and organic (developing, adaptive) understanding of the meaning and
value of the innovation in use and should also work toward a shared
language for describing the innovation and its impact.
Role of the Change Agency. If a change agency is part of a dissemination
program, the nature and quality of any linkage with intended adopter
organizations will influence the likelihood of adoption and the suc-
cess of implementation (strong indirect and moderate direct evidence).
In particular, human relations should be positive and supportive; the
two systems should have a common language, meanings, and value sys-
tems; they should share resources; the change agency should enable and
facilitate networking and collaboration among organizations; and the

Innovation in Service Organizations 613
consequences of innovations should be jointly evaluated. The change
agency should have the capacity, commitment, technical capability, com-
munication skills, and project management skills to assist with opera-
tional issues. This is particularly important in relation to technology-
based innovations, which should be disseminated as augmented products
with tools, resources, technical help, and so on (for moderate direct evi-
dence, see Lomas 2000; and Rogers 1995).
External Change Agents. Change agents employed by external agen-
cies will be more effective if they are (1) selected for their homophily and
credibility with the potential users of the innovation; (2) trained and sup-
ported to develop strong interpersonal relationships with potential users
and to explore and empathize with the user’s perspective; (3) encouraged
to communicate the users’ needs and perspective to the developers of
the innovation; and (4) able to empower the users to make independent
evaluative decisions about the innovation (for strong indirect and limited
direct evidence, see Rogers 1995).
Discussion and Recommendations
for Further Research
This study has attempted to combine a large and diverse literature into
a unifying model of the diffusion of innovations in health care orga-
nizations. Our methods were systematic and independently verifiable.
However, the literature was vast and complex, our approach was emer-
gent and somewhat unconventional, and many subjective judgments and
serendipitous discoveries were involved. A different group of researchers
setting out to answer the same research question would inevitably have
identified a different set of primary sources and made different judg-
ments about their quality and relevance. Their synthesis might have
produced a different unifying model. This is, arguably, an inherent char-
acteristic of any systematic review that addresses complex interventions
and seeks to unpack the nuances of their implementation in different
social, organizational, or environmental contexts. In this respect, a meta-
narrative review can be thought of as a particular application of a realist
review, in which the reviewer’s interpretive judgments are integral to
the synthesis process and can never be fully rationalized or standardized
(Greenhalgh et al. 2005b; Pawson et al. 2005). The findings presented
here, and especially the model in Figure 3, should therefore be seen as
“illuminating the problem and raising areas to consider” rather than

614 Trisha Greenhalgh et al.
“providing the definitive answers.” A recently published review of dif-
fusion of innovations aimed at changing individual clinician behavior,
not available when we were developing our model, was consistent with
our own conclusions (Fleuren, Wiefferink, and Paulussen 2004).
Our review affirmed many well-described themes in the literature,
such as the useful list of innovation attributes that predict (but do not
guarantee) successful adoption; the importance of social influence and the
networks through which it operates; the complex and contingent nature
of the adoption process; the characteristics (both “hard” and “soft”) of
organizations that encourage and inhibit innovation; and the messy, stop-
start, and difficult-to-research process of assimilation and routinization.
We also exposed some demons in this literature, such as the lack of em-
pirical evidence for the widely cited “adopter traits”; the focus on innova-
tions that arise centrally and are disseminated through official channels
at the expense of those that arise peripherally and spread informally; the
limited generalizability of the empirical work on product-based inno-
vation in companies to process innovation in service organizations; and
the near absence of studies focusing primarily on the sustainability of
complex service innovations.
The components of this model do not, of course, represent a com-
prehensive list of the determinants of organizational innovativeness and
successful assimilation. They are simply the areas on which research
has been undertaken and findings have been published. Conspicuously
absent from most empirical work in the service sector, for example,
is the important issue of internal politics (e.g., doctor-manager power
balances), which was identified in a single qualitative study as one of sev-
eral critical influences (Champagne et al. 1991). In an evaluation of five
projects to implement complex service innovations in primary health
care, our own team found that power relations (especially between a
project steering group and the main project worker) were critical to suc-
cessful implementation but that they were extremely diff icult to explore
systematically and raised ethical issues for the research team (Hughes
et al. 2002).
A striking finding of this extensive review was the tiny proportion
of empirical studies that acknowledged, let alone explicitly set out to
study, the complexities of spreading and sustaining innovation in ser-
vice organizations. Most studies concentrated on a few of the components
depicted in our model and failed to take account of their different in-
teractions and contextual and contingent features. This, of course, is an

Innovation in Service Organizations 615
inherent limitation of any experimental or quasi-experimental research:
The shifting baseline of context and the multiplicity of confounding
variables must be stripped away (“controlled for”) to make the research
objective (Pawson et al. 2005).
But herein lies a paradox. Context and “confounders” lie at the very
heart of the diffusion, dissemination, and implementation of complex
innovations. They are not extraneous to the object of study; they are an
integral part of it. The multiple (and often unpredictable) interactions
that arise in particular contexts and settings are precisely what deter-
mine the success or failure of a dissemination initiative. Champions, for
example, emerged in our review as a key determinant of organizational
innovation, but no amount of empirical research will provide a simple
recipe for how champions should behave that is independent of the nature
of the innovation, the organizational setting, the sociopolitical context,
and so on.
Based on the findings of this review, on some of the methodological
recommendations made by others (Green 2001; Pawson et al. 2005;
Rootman et al. 2001), and on feedback from policymakers who read
drafts of this review, we suggest that the next generation of research on
diffusion of health service innovations should be
•Theory-driven: Empirical studies should explore an explicit hypoth-
ecated link between an intervention or program and a defined out-
come. Specifically, researchers should refine their understanding
of the mechanism by which the determinants produce (or fail to
produce) the outcome of interest in a particular context.
•Process rather than “package” oriented: Researchers should avoid ques-
tions framed in terms of causal inferences, such as “Does program X
work?” or “Does strategy Y have this effect?” Rather, research ques-
tions should be framed so as to illuminate a process; for example,
“What features account for the success of program X in this context
and the failure of a comparable program in a different context?”
•Ecological: Research should recognize the reciprocal interaction be-
tween the program that is the explicit focus of research and the wider
setting in which it takes place. The latter provides a dynamic, shift-
ing baseline against which any program-related activity will occur;
each influences the other. Program-setting interactions form an
important element of data and are a particularly rich source of new
hypotheses about mechanisms of success or failure.

616 Trisha Greenhalgh et al.
•Addressed using common definitions, measures, and tools: Empirical work
should adopt standardized approaches to measuring key variables
and confounders (e.g., quality of life, implementation success) to
enable valid comparisons across studies.
•Collaborative and coordinated: Research teams should prioritize and
study research questions across multiple programs in a variety of
contexts, rather than small isolated teams “doing their own thing.”
In this way, the impact of place, setting, and context can be system-
atically studied.
•Multidisciplinary and multimethod: Research should recognize the in-
herent limitations of experimental approaches to researching open
systems and embrace a broad range of research methods emphasizing
interpretive approaches.
•Meticulously detailed: Studies should document the unique aspects
of different programs and their respective contexts and settings to
allow for meaningful comparisons across programs. Such detailed
descriptions, perhaps stored centrally as electronic appendices to
published papers and reports, could be used by future research teams
to interpret idiosyncratic findings and test rival hypotheses about
mechanisms.
•Participatory: Because of the reciprocal interactions between context
and program success, researchers should engage “on-the-ground”
service practitioners as partners in the research process. Locally
owned and driven programs produce more useful research questions
and data that are more valid for practitioners and policymakers.
When carrying out this review, we were struck by the duplication of
empirical studies and also by the number of studies that had been under-
taken without a comprehensive review of the existing relevant research,
many of which asked what appeared to be obsolete questions. Bearing in
mind our general recommendation for a more “whole-systems” approach
to researching organizational innovation, we next highlight specific areas
in which we believe further empirical research is—and, equally impor-
tant, is not—needed.
Innovations
Further research into the attributes of innovations that promote their
adoption is probably not needed. Instead, research in this area should be
directed at the following questions:

Innovation in Service Organizations 617
How do innovations in health service organizations arise, and in what
circumstances? What mix of what factors tends to produce “adoptable”
innovations (e.g., ones that have clear advantages beyond their source or-
ganization and low implementation complexity and are readily adaptable
to new contexts)?
How can innovations in health service organizations be adapted to be
perceived as more advantageous, more compatible with prevailing norms
and values, less complex, more trialable, with more observable results,
and with greater scope for local reinvention? Is there a role for a central
agency, resource center, or officially sanctioned demonstration programs
in this?
How are innovations arising as “good ideas” in local health care systems
reinvented as they are transmitted through individual and organizational
networks, and how can this process be supported or enhanced?
How can we identify “bad ideas” likely to spread so that we can
intervene to prevent this?
Adopters and Adoption
We do not recommend further descriptive studies of patterns of adop-
tion by individuals. We believe the main unanswered questions are the
following:
Why and how do people (and organizations) reject an innovation after
adopting it? (In the more than 200 empirical research studies covered in
our review, only one explicitly and prospectively studied discontinuance;
see Riemer-Reiss 1999).
What are the transferable lessons from cognitive and social psychology
about the ability and tendency of individuals to adopt particular innova-
tions in particular circumstances? For example, what can we glean from
the mainstream literature about how individuals process information,
make decisions, apply heuristics, and so on? A particularly fruitful area
is likely to be the psychological literature on the interaction between
humans and computers as it applies to the adoption and assimilation of
information and communications technology (ICT) innovations in the
service sector.
Dissemination and Social Influence
We do not recommend further “intervention” trials of the use of opinion
leaders to change the behavior of potential adopters. We already know

618 Trisha Greenhalgh et al.
from published research that opinion leadership is a complex and deli-
cate process, and research that fails to capture these process elements is
unlikely to add to what we already know. We recommend that research
into dissemination address the following questions:
What is the nature of interpersonal influence and opinion leadership
in the range of different professional and managerial groups in the health
service, especially in relation to organizational innovations? In particular,
how are key players identified and influenced?
What is the nature and extent of the social networks of different
players in the health service (both clinical and nonclinical)? How do
these networks serve as channels for social influence and the reinvention
and embedding of complex service innovations?
Who are the individuals who act as champions for organizational
innovations in health services? What is the nature of their role, and how
might it be enabled and enhanced?
Who are the individuals who act as boundary spanners among health
service organizations, especially in relation to complex service innova-
tions? What is the nature of their role, and how might it be enabled and
enhanced?
The Organizational Context
We do not recommend further survey-based research to identify struc-
tural determinants of innovativeness in health care organizations, since
the small but significant effect of structural determinants is well es-
tablished. We suggest the following questions as possible directions for
further research:
To what extent do “restructuring” initiatives (popular in health ser-
vice organizations) improve their ability to adopt, implement, and sus-
tain innovations? In particular, will a planned move from a traditional
hierarchical structure to one based on semiautonomous teams with in-
dependent decision-making power improve innovativeness?
How can we improve the absorptive capacity of service organizations
for new knowledge? In particular, what is the detailed process by which
ideas are captured from outside, circulated internally, adapted, reframed,
implemented, and routinized in a service organization, and how might
this process be systematically enhanced?
How can leaders of service organizations set about achieving a recep-
tive context for change; that is, the kind of culture and climate that

Innovation in Service Organizations 619
supports and enables change in general? A systematic review centering
on the mainstream change management literature (which we explicitly
excluded from this review) is probably the most appropriate first step
for this question.
What is the process leading to long-term routinization (with appropri-
ate adaptation and development) of innovations in health service delivery
and organization?
System Readiness for Innovation
There is relatively little systematic research on the development of sys-
tem readiness (i.e., the steps that organizations can take to assess and
anticipate the impact of an innovation). The following questions should
be addressed:
What steps must be taken by service organizations when moving
toward a state of “readiness” (i.e., with all players on board and with pro-
tected time and funding), and how can this overall process be supported
and enhanced? In particular, (1) How can tension for change be engen-
dered? (2) How can innovation-system fit best be assessed? (3) How can
the implications of the innovation be assessed and fed into the decision-
making process? (4) What measures enhance the success of efforts to
secure funding for the innovation in the resource allocation cycle? and
(5) How can the organization’s capacity to evaluate the impact of the
innovation be enhanced?
What are the characteristics of organizations that successfully avoid
taking up “bad ideas”? Are they just lucky, or do they have better mech-
anisms for evaluating the ideas and anticipating the subsequent effects?
The Outer Context
Aside from questions in the fields of political science and macro-
economics, the main research questions on the environmental context
are the following:
What is the nature of informal interorganizational networking in
different areas of activity, and how can this be enhanced through explicit
knowledge management activities (such as the appointment and support
of knowledge workers and boundary spanners)?
What is (or could be) the role of professional organizations and infor-
mal interprofessional networks in spreading innovation among health
care organizations?

620 Trisha Greenhalgh et al.
What is the cost-effectiveness of structured health care quality collab-
oratives and comparable models of quality improvement, and how can
this be enhanced? To what sort of projects in what sort of contexts should
resources for such interorganizational collaboratives be allocated?
What are the harmful effects of an external “push” (such as a policy
directive or incentive) for a particular innovation when the system is not
ready? What are the characteristics of more successful external pushes
promoting the assimilation and implementation of innovations by health
service organizations?
Implementation
Overall, we found that empirical studies of implementing and maintain-
ing innovations in service organizations (1) had been undertaken from
a pragmatic rather than an academic perspective and been presented as
“gray literature” reports (which for practical reasons we did not include
in this review); (2) were difficult to disentangle from the literature on
change management in general; and (3) were impoverished by lack of
process information. We recommend that further research focus on the
following two questions:
By what processes are particular innovations in health service delivery
and organization implemented and sustained (or not) in particular con-
texts and settings, and can these processes be enhanced? This question,
which was probably the most serious gap in the literature we uncovered
for this review, would benefit from in-depth mixed-methodology studies
aimed at building up a rich picture of process and impact.
Are there any additional lessons from the mainstream change man-
agement literature (to add to the diffusion of innovations literature re-
viewed here) for implementing and sustaining innovations in health care
organizations?
Conclusion
We believe that this extensive systematic review has produced three
important outputs: (1) a parsimonious and evidence-based model for
considering the diffusion of innovations in health service organizations;
(2) clear knowledge gaps on which further research on the diffusion
of innovations in service organizations should be focused; and (3) a ro-
bust and transferable methodology for systematically reviewing complex

Innovation in Service Organizations 621
research evidence (methodological issues, especially the transferability of
this method to other systematic review topics, are discussed in more de-
tail in Greenhalgh et al. 2005b). Research commissioners, in particular,
should note our suggestions for where further research is unlikely to add
substantially to the knowledge base. We encourage other research teams
to test both our proposed model for the diffusion of service innovations
and our proposed methodology for a systematic review of diffuse bodies
of literature.
References
Abrahamson, E. 1991. Managerial Fads and Fashions: The Diffusion and
Rejection of Innovation. California Management Review 16:586–612.
Abrahamson, E., and G. Fairchild. 1999. Management Fashion: Life-
cycles, Triggers and Collective Learning Processes. Administrative
Sciences Quarterly 44(4):708–40.
Adler, P.S., S-W. Kwon, and J.M.K. Singer. 2003. The “Six-
West” Problem: Professionals and the Intraorganizational Dif-
fusion of Innovations, with Particular Reference to the Case
of Hospitals. Working paper 3–15. Marshall School of Busi-
ness, University of Southern California. Full text available
on http://www.marshall.usc.edu/web/MOR.cfm?doc id=5561, Los
Angeles.
Anderson, N.R., and N.A.West. 1998. Measuring Climate for Work
Group Innovation. Journal of Organizational Behaviour 19:235–58.
Aubert, B.A., and G. Hamel. 2001. Adoption of Smart Cards in the
Medical Sector: The Canadian Experience. Social Science & Medicine
53(7):879–94.
Backer, T.E., and E.M. Rogers. 1998. Diffusion of Innovations The-
ory and Work-Site AIDS Programs. Journal of Health Communication
3(1):17–28.
Baldridge, J.V., and R.A. Burnham. 1975. Organisational Innovation:
Individual Organisational and Environmental Impacts. Administra-
tive Sciences Quarterly 20:165–76.
Barnsley, J., L. Lemieux-Charles, and M.M. McKinney. 1998. Integrat-
ing Learning into Integrated Delivery Systems. Health Care Manage-
ment Review 23(1):18–28.
Bass, F.M. 1969. A New Product Growth Model for Consumer Durables.
Management Science 13(5):215–27.
Bate, S.P. 2004. The Role of Stories and Storytelling in Organisational
Change Efforts: A Field Study of an Emerging “Community of

622 Trisha Greenhalgh et al.
Practice” within the UK National Health Service. In Narrative Re-
search in Health and Illness, edited by B. Hurwitz, T. Greenhalgh,
and V. Skultans. Oxford: Blackwell.
Becker, M.H. 1970. Factors Affecting the Dif fusion of Innovation among
Health Professionals. American Journal of Public Health 60:294–
304.
Bourdenave, J.D. 1976. Communication of Agricultural Innovations in
Latin America: The Need for New Models. Communication Research
3(2):135–54.
Buckler, S.A., and C. Zein. 1996. The Spirituality of Innovation: Learn-
ing from Stories. Journal of Product Innovation Management, September
13, 391–405.
Burns, L.R., and D.R. Wholey. 1993. Adoption and Abandonment of
Matrix Management Programs: Effects of Organizational Charac-
teristics and Interorganizational Networks. Academy of Management
Journal 36:106–38.
Burt, R.S. 1973. The Differential Impact of Social Integration on Par-
ticipation in the Diffusion of Innovations. Social Science Research
2(2):125–44.
Castle, N.G. 2001. Innovation in Nursing Homes: Which Facilities Are
the Early Adopters? Gerontologist 41(2):161–72.
Champagne, F., J-L. Denis, R. Pineault, and A. Contandriopoulos. 1991.
Structural and Political Models of Analysis of the Introduction of
an Innovation in Organizations: The Case of the Change in the
Method of Payment of Physicians in Long-Term Care Hospitals.
Health Services Management Research 4(42):94–111.
Coleman, J.S., E. Katz, and H. Menzel. 1966. Medical Innovations: A
Diffusion Study. New York: Bobbs-Merrill.
Czarniawska, B. 1998. A Narrative Approach to Organization Studies. Qual-
itative Research Methods Series 43. London: Sage.
Damanpour, F. 1991. Organisational Innovations: A Meta-Analysis of
Effects of Determinants and Moderators. Academy of Management
Journal 34:555–90.
Damanpour, F. 1992. Organizational Size and Innovation. Organization
Studies 13(3):375–402.
Damanpour, F. 1996. Organizational Complexity and Innovation: De-
veloping and Testing Multiple Contingency Models. Management
Science 42:693–716.
Dearing, J.W., and O. And. 1994. Portraying the New: Communication
between University Innovators and Potential Users. Science Commu-
nication 16(1):11–42.
Denis, J.L., Y. Hebert, A. Langley, D. Lozeau, and L.H. Trottier. 2002.
Explaining Diffusion Patterns for Complex Health Care Innova-
tions. Health Care Management Review 27(3):60–73.

Innovation in Service Organizations 623
Dirksen, C.D., A.J. Ament, and P.M. Go. 1996. Diffusion of Six Sur-
gical Endoscopic Procedures in the Netherlands. Stimulating and
Restraining Factors. Health Policy 37(2):91–104.
Dixon-Woods, M., S. Agarwal, D. Jones, B. Young, and A. Sutton. 2004.
Synthesising Qualitative and Quantitative Evidence: A Review of
Methods. Journal of Health Services Research & Policy,inpress.
Dobbins, M., R. Cockerill, and J. Barnsley. 2001. Factors Affecting the
Utilization of Systematic Reviews. International Journal of Technology
Assessment in Health Care 17:203–14.
Dopson, S., L. Fitzgerald, E. Ferlie, J. Gabbay, and L. Locock. 2002. No
Magic Targets. Changing Clinical Practice to Become More Evidence
Based. Health Care Management Review 37:35–47.
Dufault, M.A., C. Bielecki, E. Collins, and C. Willey. 1995. Changing
Nurses’ Pain Assessment Practice: A Collaborative Research Uti-
lization Approach. Journal of Advanced Nursing 21(4):634–45.
Edmondson, A.C., R.M. Bohmer, and G.P. Pisano. 2001. Disrupted
Routines: Team Learning and New Technology Implementation in
Hospitals. Administrative Sciences Quarterly 46(4):685–716.
Elliott, S.J., S.M. Taylor, R. Cameron, and R. Schabas. 1998. Assessing
Public Health Capacity to Support Community-Based Heart Health
Promotion: The Canadian Heart Health Initiative, Ontario Project
(CHHIOP). Health Education Research 13(4):607–22.
Eveland, J.D. 1986. Diffusion, Technology Transfer and Implementa-
tion. Knowledge: Creation, Diffusion, Utilisation 8(2):303–22.
Exworthy, M., L. Berney, and M. Powell. 2003. How Great Expectations
in Westminster May Be Dashed Locally: The Local Implementation
of National Policy on Health Inequalities. Policy & Politics 30(1):79–
96.
Fennell, M.L., and R.B. Warnecke. 1988. The Diffusion of Medical Inno-
vations: An Applied Network Analysis. New York: Plenum.
Ferlie, E., J. Gabbay, L. Fitzgerald, L. Locock, and S. Dopson. 2001.
Evidence-Based Medicine and Organisational Change: An Overview
of Some Recent Qualitative Research. In Organisational Behaviour and
Organisational Studies in Health Care: Reflections on the Future, edited
by L. Ashburner. Basingstoke: Palgrave.
Fitzgerald, L., E. Ferlie, M. Wood, and C. Hawkins. 2002. Interlocking
Interactions, the Diffusion of Innovations in Health Care. Human
Relations 55(12):1429–49.
Flamm, B.L., D.M. Berwick, and A. Kabcenell. 1998. Reducing Ce-
sarean Section Rates Safely: Lessons from a “Breakthrough Series”
Collaborative. Birth 25:117–24.
Fleuren, M., K. Wiefferink, and T. Paulussen. 2004. Determinants of In-
novation within Health Care Organizations: Literature Review and
Delphi Study. International Journal of Quality Health Care 16:107–23.

624 Trisha Greenhalgh et al.
Fonseca, J. 2001. Complexity and Innovation in Organisations. London:
Routledge.
Foy, R., G. MacLennan, J. Grimshaw, G. Penney, M. Campbell, and R.
Grol. 2002. Attributes of Clinical Recommendations That Influence
Change in Practice Following Audit and Feedback. Journal of Clinical
Epidemiology 55(7):717–22.
Gabriel, Y. 2000. Storytelling in Organisations: Facts, Fictions and Fantasies.
Oxford: Oxford University Press.
Gladwin, J., R.A. Dixon, and T.D. Wilson. 2002. Rejection of an Inno-
vation: Health Information Management Training Materials in East
Africa. Health Policy & Planning 17(4):354–61.
Goes, J.B., and S.H. Park. 1997. Interorganisational Links and Innova-
tion: The Case of Hospital Services. Academy of Management Journal
40(3):673–96.
Gosling, A.S., J.I. Westbrook, and J. Braithwaite. 2003. Clinical Team
Functioning and IT Innovation: A Study of the Diffusion of a Point-
of-Care Online Evidence System. Journal of the American Medical
Information Association 10(3):244–51.
Granados, A., E. Jonsson, H.D. Banta, L. Bero, A. Bonair, C. Cochet,
N. Freemantle, R. Grilli, J. Grimshaw, E. Harvey, R. Levi,
D. Marshall, A. Oxman, L. Pasart, V. Raisanen, E. Rius, and J.A.
Espinas. 1997. EUR-ASSESS Project Subgroup Report on Dissem-
ination and Impact. International Journal of Technology Assessment in
Health Care 13(2):220–86.
Granovetter, M., and R. Soong. 1983. Threshold Models of Diffusion
and Collective Behavior. Journal of Mathematical Sociology 9:165–79.
Green, L.W. 2001. From Research to “Best Practices” in Other Set-
tings and Populations. American Journal of Health Behavior 25:165–
78.
Green, P.L. 1998. Improving Clinical Effectiveness in an Integrated Care
Delivery System. Journal for Healthcare Quality 20(6):4–8.
Greenhalgh, T., G. Robert, P. Bate, O. Kyriakidou, F. Macfarlane, and
R. Peacock. 2005a (forthcoming). Diffusion of Innovations in Health
Service Organisations: A Systematic Literature Review. Oxford: Black-
well.
Greenhalgh, T., G. Robert, F. Macfarlane, P. Bate, O. Kyriakidou, and
R. Peacock. 2005b. Storylines of Research: A Meta-Narrative Per-
spective on Systematic Review. Social Science and Medicine.Inpress.
Grilli, R., and J. Lomas. 1994. Evaluating the Message: The Relationship
between Compliance Rate and the Subject of a Practice Guideline.
Medical Care 32(3):202–13.
Grimshaw, J.M., R.E. Thomas, G. MacLennan, C. Fraser, C.R. Ramsay,
L. Vale, P. Whitty, M.P. Eccles, L. Matowe, L. Shirran,

Innovation in Service Organizations 625
M. Wensing, R. Dikstra, C. Donaldson, and A. Hutchinson. 2004.
Effectiveness and Ef ficiency of Guideline Dissemination and Imple-
mentation Strategies. Health Technology Assessment Report 8(6):1–72.
Gustafson, D.H., F. Sainfort, M. Eichler, L. Adams, M. Bisognano, and
H. Steudel. 2003. Developing and Testing a Model to Predict Out-
comes of Organizational Change. Health Services Research 38(2):751–
76.
Hall, G.E., and S.M. Hord. 1987. Change in Schools. Albany: State Uni-
versity of New York Press.
Horbar, J.D., J. Rogowski, P.E. Plsek, P. Delmore, W.H. Edwards,
J. Hocker, A.D. Kantak, P. Lewallen, W. Lewis, E. Lewit, C.J.
McCarroll, D. Mujsce, N.R. Payne, P. Shiono, R.F. Soll, K. Leahy,
and J.H. Carpenter. 2001. Collaborative Quality Improvement for
Neonatal Intensive Care. Pediatrics 107(1):14–22.
House, R., D.M. Rousseau, and M. Thomas-Hunt. 1995. The Meso
Paradigm: A Framework for the Integration of Micro and Macro Or-
ganisational Behaviour. Research in Organizational Behaviour 17:71–
114.
Hughes, J., C. Humphrey, S. Rogers, and T. Greenhalgh. 2002. Evidence
into Action: Changing Practice in Primary Care. Occasional Papers,
Royal College of General Practitioners 84:i–51.
Kanter, R.M. 1988. When a Thousand Flowers Bloom: Structural, Col-
lective and Social Conditions for Innovation in Organisation. In
Research in Organisational Behaviour, edited by B.M. Staw and L.L.
Cummings. Greenwich, Conn.: JAI Press.
Kimberly, J.R., and J.M. Evanisko. 1981. Organisational Innovation:
The Influence of Individual, Organisational and Contextual Factors
on Hospital Adoption of Technological and Administrative Innova-
tion. Academy of Management Journal 24(4):689–713.
Kitson, A., G. Harney, and B. McCormack. 1998. Enabling the Imple-
mentation of Evidence Based Practice: A Conceptual Framework.
Quality in Health Care 7:149–58.
Kuhn,T.S. 1962. The Structure of Scientif ic Revolutions. Chicago: University
of Chicago Press.
Leape, L.L., A.I. Kabcenell, T.K. Gandhi, P. Carver, T.W. Nolan, and
D.M. Berwick. 2000. Reducing Adverse Drug Events: Lessons from
a Breakthrough Series Collaborative. Joint Commission Journal on
Quality Improvement 26:321–31.
Locock, L., S. Dopson, D. Chambers, and J. Gabbay. 2001. Understand-
ing the Role of Opinion Leaders in Improving Clinical Effectiveness.
Social Science & Medicine 53:745–57.
Lomas, J. 2000. Using “Linkage and Exchange” to Move Research into
Policy at a Canadian Foundation. Health Affairs 19(3):236–40.

626 Trisha Greenhalgh et al.
Markham, S.K. 1998. A Longitudinal Examination of How Champi-
ons Influence Others to Support Their Projects. Journal of Product
Innovation Management 15(6):490–504.
Marshall, J.G. 1990. Diffusion of Innovation Theory and End-User
Searching. Library & Information Science Research 6(1):55–69.
McCormick, L.K., A.B. Steckler, and K.R. Mcleroy. 1995. Diffusion of
Innovations in Schools: A Study of Adoption and Implementation
of School-Based Tobacco Prevention Curricula. American Journal of
Health Promotion 9(3):210–9.
Meyer, A.D., and J.B. Goes. 1988. Organisational Assimilation of Inno-
vations: A Multi-Level Contextual Analysis. Academy of Management
Review 31:897–923.
Meyer, M., D. Johnson, and C. Ethington. 1997. Contrasting Attributes
of Preventive Health Innovations. Journal of Communication 47:112–
31.
Meyers, P.W., K. Sivakumar, and C. Nakata. 1999. Implementation
of Industrial Process Innovations: Factors, Effects, and Marketing
Implications. Journal of Product Innovation Management 16(3):295–
311.
Newton, J., J. Graham, K. McLoughlin, and A. Moore. 2003. Recep-
tivity to Change in a General Medical Practice. British Journal of
Management 14:143–53.
Nonaka, I., and H. Takeuchi. 1995. The Knowledge Creation Company:
How Japanese Companies Create the Dynamics of Innovation. New York:
Oxford University Press.
Nystrom, P.C., K. Ramamurthy, and A.L. Wilson. 2002. Organizational
Context, Climate and Innovativeness: Adoption of Imaging Technol-
ogy. Journal of Engineering and Technology Management 19(3–4):221–
47.
O’Connor, G.T., S.K. Plume, E.M. Olmstead, J.R. Morton, C.T.
Maloney, W.C. Nugent, F. Hernandez, Jr., R. Clough, B.J. Leavitt,
L.H. Coffin, C.A. Marrin, D. Wennberg, J.D. Birkmeyer, D.C.
Charlesworth, D.J. Malenka, H.B. Quinton, and J.F. Kasper. 1996.
A Regional Intervention to Improve the Hospital Mortality Asso-
ciated with Coronary Artery Bypass Graft Surgery. The Northern
New England Cardiovascular Disease Study Group. Journal of the
American Medical Association 275:841–6.
O’Neill, H.M., P.W. Pouder, and A.K. Buchholtz. 2002. Patterns in the
Diffusion of Strategies across Organisations: Insights from the Inno-
vation Diffusion Literature. Academy of Management Review 23(1):98–
114.
Øvretveit, J. 2003. Reviewing Medical Management Research for Decision-
Makers: Methodological Issues in Carrying Out Systematic Reviews of Med-
ical Management Research. Stockholm: Karolinska Institute (Medical
Management Centre internal discussion document).

Innovation in Service Organizations 627
Øvretveit, J., P. Bate, P. Cleary, S. Cretin, D. Gustafson, K. McInnes, H.
McLeod, T. Molfenter, P. Plsek, G. Robert, S. Shortell, and T. Wilson.
2002. Quality Collaboratives: Lessons from Research. Quality and
Safety in Health Care 11:345–51.
Patel, V. 1996. Cognition and Technology in Health Education Research.
Canadian Journal of Public Health 87 (suppl. 2):S63–7.
Pawson, R., T. Greenhalgh, G. Harvey, and K. Walshe. 2005 (forth-
coming). Using Realist Methods to Produce Syntheses of Evidence
for Use by Managers and Policy Makers. London: Canadian Health
Services Research Foundation and NHS SDO Programme.
Pettigrew, A.M., and L. McKee. 1992. Shaping Strategic Change. Making
Change in Large Organisations. London: Sage.
Plsek, P. 2003. Complexity and the Adoption of Innovation in
Health Care. Paper presented at Accelerating Quality Improve-
ment in Health Care: Strategies to Accelerate the Diffusion of
Evidence-Based Innovations. Washington, D.C.: National Institute
for Healthcare Management Foundation and National Committee
for Quality in Health Care.
Potvin, L., S. Haddad, and K.L. Frohlich. 2001. Beyond Process and
Outcome Evaluation: A Comprehensive Approach for Evaluat-
ing Health Promotion Programmes. WHO Regional Publications,
European Series 92:45–62.
Rashman, L., and J. Hartley. 2002. Leading and Learning? Knowl-
edge Transfer in the Beacon Council Scheme. Public Administration
80(3):523–42.
Riemer-Reiss, M.L. 1999. Applying Rogers’ Diffusion of Innovations
Theory to Assistive Technology Discontinuance. Journal of Applied
Rehabilitation Counseling 30(4):16–21.
Robertson, T., and Y. Wind. 1983. Organisational Cosmopolitanism and
Innovation. Academy of Management Journal 26:332–8.
Rogers, E.M. 1995. Diffusion of Innovations. New York: Free Press.
Rogers, E.M., and D.L. Kincaid. 1981. Communication Networks: Toward
a New Paradigm for Research. New York: Free Press.
Rogowski, J.A., J.D. Horbar, P.E. Plsek, B.L. Schuurmann, J. Deterding,
W.H. Edwards, J. Hocker, A.D. Kantak, P. Lewallen, W. Lewis,
E. Lewit, C.J. McCarroll, D. Mujsce, N.R. Payne, P. Shiono, R.F.
Soll, and K. Leahy. 2001. Economic Implications of Neonatal In-
tensive Care Unit Collaborative Quality Improvement. Pediatrics
107(1):23–9.
Rootman, I., M. Goodstadt, L. Potvin, and J. Springett. 2001. A Frame-
work for Health Promotion Evaluation. WHO Regional Publica-
tions, European Series 92:7–38.
Schon, D.A. 1963. Champions for Radical New Inventions. Harvard
Business Review 41:77–86.

628 Trisha Greenhalgh et al.
Shane, S. 1995. Uncertainty Avoidance and the Preference for Innovation
Championing Roles. Journal of International Business Studies 26(1):47–
68.
Taylor, S.M., S. Elliott, K. Robinson, and S. Taylor. 1998. Community-
Based Heart Health Promotion: Perceptions of Facilitators and Bar-
riers. Canadian Journal of Public Health/Revue canadienne de sant´e
publique 89(6):406–9.
Thomson O’Brien, M.A., A.D. Oxman, D.A. Davis, R.B. Haynes, and N.
Freemantle. 2003. Local Opinion Leaders. Cochrane Database Systems
Review 1.
Timmons, S. 2001. How Does Professional Culture Influence the Success
or Failure of IT Implementation in Health Services? In Organisational
Behaviour and Organisational Studies in Health Care: Reflections on the
Future, edited by L. Ashburner. Basingstoke: Palgrave.
Tornatsky, L.G., and K.J. Klein. 1982. Innovation Characteristics and
Innovation-Adoption-Implementation: A Meta-Analysis of Find-
ings. Transactions on Engineering Management 29(1):28–45.
Tushman, M. 1977. Special Boundary Roles in the Innovation Process.
Administrative Sciences Quarterly 22:587–605.
UK Department of Health. 2001. The NHS Plan. London: NHS
Executive.
Valente, T.W. 1995. Network Models of the Diffusion of Innovations.
Cresskill, N.J.: Hampton.
Valente, T.W. 1996. Social Network Thresholds in the Diffusion of
Innovations. Social Networks 18(1):69–89.
VandeVen, A.H., D.E. Polley, R. Garud, and S. Venkataraman. 1999.
The Innovation Journey. Oxford: Oxford University Press.
Wejnert, B. 2002. Integrating Models of Diffusion of Innovations: A
Conceptual Framework. Annual Review of Sociology 28:297–326.
West, E., D.N. Barron, J. Dowsett, and J.N. Newton. 1999. Hierarchies
and Cliques in the Social Networks of Health Care Professionals:
Implications for the Design of Dissemination Strategies. Social Science
& Medicine 48(5):633–46.
Westphal, J.D., R. Gulati, and S.M. Shortell. 1997. Customization or
Conformity? An Institutional and Network Perspective on the Con-
tent and Consequences of Total Quality Management Adoption.
Administrative Sciences Quarterly 42(366):394.
Wilson, A.L., K. Ramamurthy, and P.C. Nystrom. 1999. A Multi-
Attribute Measure for Innovation Adoption: The Context of Imaging
Technology. Transactions on Engineering Management 46(3):311–21.
Wolfe, B. 1994. Organisational Innovation: Review, Critique and Sug-
gested Research Directions. Journal of Management Studies 31:405–
31.

Innovation in Service Organizations 629
Yetton, P., R. Sharma, and G. Southon. 1999. Successful IS Innovation:
The Contingent Contributions of Innovation Characteristics and
Implementation Process. Journal of Information Technology 14(1):53–
68.
Zahra, A.S., and G. George. 2002. Absorptive Capacity: A Review,
Reconceptualization and Extension. Academy of Management Review
27(2):185–203.
Acknowledgments: This study was funded by the UK Department of Health.
Thanks to the many people, too numerous to mention individually, who pro-
vided formative feedback on our work during this project or comments on drafts
of the final report.