Intervening to change health system performance for the better
It is temptingly easy to treat improvement interventions as if they are drugs—technical,
stable and uninfluenced by the environment in which they work. Doing so makes life
so much easier for everyone. It allows improvement practitioners to plan their work
with a high degree of certainty, funders to be confident that they know what they
are buying and evaluators to focus on what really matters—whether or not ‘it’ works.
But of course most people know that life is not as simple as that. Experienced improvers
have long recognised that interventions—the specific tools and activities introduced
into a healthcare system with the aim of changing its performance for the better1—flex
and morph. Clever improvers watch and describe how this happens. Even more clever
improvers plan and actively manage the process in a way that optimises the impact
of the improvement initiative.
The challenge is that while most improvers (the authors included) appreciate the importance
of carefully designing an improvement intervention, they (we) rarely do so in a sufficiently
clever way. In this article, we describe our attempts as an experienced team of practitioners,
improvers, commissioners and evaluators to design an effective intervention to improve
the safety of people living in care homes in England. We highlight how the design
of the intervention, as described in the original grant proposal, changed significantly
throughout the initiative. We outline how the changes that were made resulted in a
more effective intervention but how our failure to design a better intervention from
the start reduced the overall impact of the project. Drawing on the rapidly expanding
literature in the field and our own experience, we reflect on what we would do differently
if we could have our time again.
A practical case study—an initiative to improve the safety of people living in care
homes
A growing number of vulnerable older people are living in care homes and are at increased
risk of preventable harm. We carried out a safety improvement programme with a linked
participatory multimethod evaluation2 in care homes in the south east of England.
Ninety homes were recruited in four separate cohorts over a 2-year period. Our aim
was to reduce the prevalence of three of the most common safety events in the sector—falls,
pressure ulcers and urinary tract infections—and thereby to reduce unnecessary attendances
at emergency departments and admissions to hospital.
In the original proposal submitted to the funding body, we described a multifaceted
intervention comprising three main elements:
The measurement and benchmarking of (i) the prevalence of the target safety incidents
using a nationally designed tool called the NHS Safety Thermometer3 and (ii) rates
of emergency department attendances and hospital admissions using routinely collected
data.
Training in quality improvement methods provided initially by a team of NHS improvement
advisors and then, using a ‘train the trainer’ model, by practitioners working with
or in the care homes.
The use of a specially adapted version of the Manchester Patient Safety Framework,4
(Marshall M, de Silva D, Cruickshank L, et al. Understanding the safety culture of
care homes; insights from the adaptation of a health service safety culture assessment
tool for use in the care home sector (submitted to BMJ Qual Saf, August 2016), a formative
assessment tool which provides insights into safety culture for frontline teams.
The intervention was underpinned by a strong emphasis on support and shared learning
using communities of practice and online resources facilitated by the improvement
team.
The programme theory hypothesised that the three main elements of the intervention
(benchmarking, learning improvement skills and cultural awareness) would reduce the
prevalence of safety events, that this would lead to a reduction in emergency department
attendances and hospital admissions and that both outcomes would reduce system costs
as well as improving the quality of care for residents. The intervention was co-designed
by improvement researchers in the evaluation team, the improvement team in the local
government body responsible for commissioning care home services and a senior manager
of one of the participating care homes. The design was influenced by a combination
of theory, published empirical evidence and the personal knowledge and experience
of the commissioners and care home manager.
We built in a 6-month preparatory period at the start of the programme, prior to implementing
the intervention with the first cohort of care homes. This period was used to recruit
staff, establish the project infrastructure and build relationships between the care
homes and the improvement and evaluation teams. Only when the programme formally started
did we begin to expose some of the deficiencies in the planned intervention. Table
1 describes the different components of the intervention, whether it was part of the
original plan or introduced at a later stage, and, based on our participatory evaluation,
how it was implemented and the extent to which it was used.
Table 1
The original intervention and how it evolved
Intervention component
Original/added later
Ways in which the component were implemented
Extent to which component was used
NHS Safety Thermometer (NHS designed and owned online tool for collecting process
and outcomes data)
Original
Implemented with first cohort and offered to all of second cohort, then replaced by
Safety Cross and Monthly Mapping tools (see below)
66% of first cohort homes tried the Safety Thermometer. About one-third input data
Active involvement of staff, residents and relatives in sharing data and co-creating
improvement solutions
Original
Staff initially slow to share data but became enthusiastic as project progressed.
Residents and relatives hardly actively involved at all but project details and data
displayed on public notice boards in most homes
Fewer than 10% of first cohort homes shared Safety Thermometer data. Eighty per cent
of homes used the Safety Cross and displayed this for staff, residents and families
to see. Sixty per cent displayed graphs from the Monthly Mapping tool
Training for care home staff in improvement methodologies
Original
Quality improvement training was provided initially by the NHS staff, then adapted
and provided by the improvement team
All homes took part in training. In first cohort, this was chiefly home managers but
in subsequent cohorts some senior carers also attended
Participants able to deliver the training to peers (train-the-trainer)
Original
Formal train-the-trainer model was not implemented though local advocates (‘champions’)
were encouraged to roll out learning to others
Champions were found to work well to spread learning informally
Intervention toolkit containing a compendium of evidenced-based interventions for
each of the domains of the Safety Thermometer
Original
Toolkit with worksheets and information sheets developed
All homes received a hard copy and an online version. Unclear how much they were used
by first cohort and then dropped as Safety Thermometer replaced by Safety Cross
Safety culture assessed using the MaPSaF tool at three time points (before, during
and after PROSPER), using the tool to understand and address barriers to change
Original
MaPSaF revised and tested in different ways with various cohorts
Use not prioritised by the improvement team or by the homes. Small number of homes
actively used it. Progressively more significant changes made to the tool for each
cohort to make it more relevant
Communities of practice
Original
Three community of practice events held throughout project
Between a half and two-thirds of homes attended the events
Improvement tools and case studies uploaded to resource tool for peer learning
Original
Knowledge hub set up and documents uploaded periodically, mainly copies of things
sent by email
10% of homes signed up and none of them posted information
Ongoing support from improvement team including meetings, visits and telephone conferences
Original
Facilitators visited homes with varied frequency. During the intensive phase, some
homes were visited monthly and others every 3–4 months. Group telephone conferences
were not used
Some homes received regular support and others did not. Some homes reported that they
had no contact with their allocated improvement adviser for 6 months
‘Safety Cross' for displaying information about monthly incidents replaced Safety
Thermometer (see above)
Addition
Used from cohort two homes onwards then also rolled out to cohort one
About 80% of homes reported using it
‘Monthly Mapping tool’ using graphs with monthly data to track changes over time and
compare averages
Addition
All homes were invited to provide data about the monthly incidence of harms. From
cohort three onwards, homes were given access to an online tool
About 60% of homes provided some data. One-quarter used the tool regularly without
prompting
Provision of resources such as information posters, certificates of training, mirrors
to view pressure ulcers and other tangible resources
Addition
Resources developed ad hoc
Homes offered tools during community of practice and visits. Variable uptake depending
on focus. Resources appeared to be highly appreciated
Provision of additional training beyond improvement methods courses, such as training
in infection control and pressure ulcers
Addition
Twenty-six training sessions run
About 50% of homes participated
Coordination with partner organisations in the NHS
Addition
Varied by geographical area
Varied by geographical area
Monthly newsletter
Addition
Sent monthly to participating homes
Sixty per cent of home managers reported reading it
Green=implemented as planned; Amber=partly implemented as planned; Red=not implemented
as planned.
MaPSaF, Manchester Patient Safety Framework.
The evaluation found that four of the nine original components of the intervention
were not implemented as planned and two were only partially implemented as planned.
Only three of the nine were implemented in line with the original proposal. Five of
the six new intervention components, designed and implemented while the initiative
was taking place, were fully implemented. Qualitative evaluative data, collected using
interviews, surveys and observations, demonstrated changes in the attitudes of frontline
staff to safety and changes in their working practices. However, quantitative data
suggested only small and variable changes of questionable statistical significance
in the prevalence of safety incidents, and no impact on the background rising rates
of emergency department attendances and hospital admissions.
Success or failure?
Perhaps we should not be too hard on ourselves. On the surface at least, our intervention
was more sophisticated than that seen for most improvement projects.5 The multifaceted
intervention had complementary measurement, educational and culture-change elements
and was co-designed by a wide group of stakeholders, including a practitioner and
experienced improvement science academics. We based the design on a reasonable programme
theory and an explicit logic model. We recognised the need to adapt off-the-shelf
tools to the local context and to build in a preparatory period prior to formally
evaluating the intervention. And we purposefully chose a participatory and formative
evaluation model to support a feedback cycle as the initiative progressed.
As a project team, we thought that we had designed the original intervention thoughtfully
and carefully but the findings of our evaluation suggested that we could have done
a lot better. Reflecting towards the end of the programme, we considered a number
of possible explanations: we did not put enough time and effort into designing the
intervention; we designed a sound intervention which was not implemented sufficiently
well or was implemented without an adequate understanding of the context and our expectations
were naïve that an intervention at such an early stage of development would have a
significant impact. We then revisited the literature to examine these hypotheses.
What the literature suggests we should have done
There is no shortage of increasingly sophisticated theory, empirical evidence and
learned commentary that could have guided our design decisions. Much of the thinking
about interventions is relatively new; a state-of-the-art review of improvement published
in the Lancet more than 15 years ago made no specific reference to the ways in which
interventions morph when applied in practice.6 In contrast, more recent international
guidance on designing, conducting and writing up improvement projects highlights the
importance of describing how improvement interventions change.7 In brief, a number
of themes relating to the design of effective interventions are emerging in the literature.
First, the importance of using theory (‘a chain of reasoning’) to optimise the design
and effectiveness of interventions is highlighted.8 A commonsense rather than an overly
academic approach to theory is being advocated as a way of reducing the risk of the
‘magical thinking’, which encourages improvers to use interventions that look superficially
attractive but for which the mechanisms of action are unclear.8
9 Alongside the use of theory, there is a growing interest in the application of ‘design
thinking’ as a strategy for ensuring that the problem has been clearly identified
and a way of addressing complex problems in rapidly changing environments.10 Second,
the importance of having an explicit method, such as the Institute for Healthcare
Improvement's Model for improvement using Plan-Do-Study-Act cycles, is described and
also understanding how to use the methods to their full potential.11 Third, there
is a growing emphasis on the extent to which improvement interventions are social
as well as technical in nature, and how their effectiveness is a consequence of a
complex interaction between people, organisational structures and processes.12
13 Fourth, the literature describes how what people do (intervention), how they do
it (implementation) and the wider environment (context) are interdependent and some
people are suggesting that the traditional differentiation between this classic triad
is no longer helpful.14
Fifth, there is a growing consensus that improvement efforts are being evaluated too
early in their development and as a consequence are being judged unfairly as being
ineffective.15
16 Instead, there are calls for interventions to be categorised according to the ‘degree
of belief’ that they will work16 and how this belief becomes stronger as a project
progresses. Interventions in the early ‘innovation’ phase should be evaluated using
different methods from those in the later ‘testing’ or ‘spread’ phases. They may also
have a different intent, for example, changes in behaviour may be seen as ‘success’
before measurable changes in outcome are achieved. Sixth, drawing on the expanding
field of knowledge mobilisation,17
18 experts are calling for a more active process of co-design of improvement initiatives
involving service users, practitioners and improvers, and also academics, with all
of these stakeholders contributing to participatory models of evaluation.19
What we would do differently?
Having reviewed the literature, we came to the conclusion that each of the post hoc
hypotheses were reasonable explanations for what in the field of improvement were
not uncommon results, but were nevertheless disappointing. In future, we will put
more effort into designing the intervention from the very start. We will think through
the design issues in sufficient detail to not only persuade the funder of the project
but also to persuade ourselves that it will work in practice. We will describe a programme
theory in greater detail based on a better understanding of the contextual factors
which could impact on the feasibility and effectiveness of the initiative, and we
will use design thinking to rigorously frame the problem from the start.
We will work through in more detail and more systematically how to use current thinking
about intervention design and its applicability to our project. We will build-in a
similar or even longer preparatory period and will use that period to test and refine
the intervention. We will not rely on a single senior care home manager to provide
a practitioner view for the original proposal and we will seek a wide range of views
from frontline staff and from care home residents in an inclusive and iterative way.
We will not assume that the intervention can be implemented as described in the proposal
and we will be more sensitive to the resource constraints under which the improvement
team and the care homes are operating.
If we do all of this, the outcome will almost certainly be better.
Final reflections
Improvement initiatives are sometimes planned on the hard high ground, but they are
put into effect in the swampy lowlands.20 As we are more than aware, frontline practice
is messy. And as we have described in this paper, it is never possible to do things
perfectly and good improvers are always learning. But as the improvement movement
matures, we are getting to the stage where we could and should be doing better. It
needs to be seen as a professional rather than an amateur sport. The importance of
understanding that improvement interventions are not like drugs or medical devices,
and that flexibility needs to be built into their design and delivery, is uncontestable.
But is it no longer acceptable to use the need for flexibility as an excuse for a
lack of thought and planning. As improvement becomes more rigorous, perhaps improvement
practitioners will be able to plan their work with a higher degree of certainty, funders
will be more confident that they know what they are buying and evaluators will be
able to focus on whether and how ‘it’ works.