2. About us
We shine a light on
how to make successful
change happen
The Health Foundation is an
independent charity committed
to bringing about better health
and health care for people
in the UK.
We connect what works
on the ground with
effective policymaking
and vice versa.
3.
4. Evaluating new models of care
Analytical methods have been developed to assess the
impacts of new models of care, often (not exclusively) in
research settings
A common approach is to:
• Identify the intervention group within existing data
• Retrospectively select a ‘control’ group, e.g. by matching
to the characteristics of the intervention group
• Compare the outcomes of these groups
• Other methods are possible
5. Example: Birmingham OwnHealth
A telephone-based health coaching trial in Birmingham for
people with heart failure, CHD, COPD, diabetes and history
of hospital use. Expected to reduce hospital admissions by
giving participants personalised care plan and monthly
‘check-in’ phone calls.
• 2,698 people recruited to receive the intervention
• Retrospectively matched to ‘control group’ with similar
characteristics
• Looked at emergency hospital admissions, pre- and post-
intervention, for each group
6. Example: Birmingham OwnHealth
Steventon A, Tunkel S, Blunt I, Bardsley M. Effect of telephone health coaching (Birmingham OwnHealth) on
hospital use and associated costs: cohort study with matched controls. BMJ. 2013;347:f4585.
7. Example: Birmingham OwnHealth
Steventon A, Tunkel S, Blunt I, Bardsley M. Effect of telephone health coaching (Birmingham OwnHealth) on
hospital use and associated costs: cohort study with matched controls. BMJ. 2013;347:f4585.
8. New care models
Information on the impact of new care models on patient
outcomes and cost needed for:
• Local improvement
• National policymaking
Critical for evaluation arrangements to be ‘baked in’ from
the very start to provide timely, regular, high-quality
feedback
9. The challenge
We have the methods, but need to develop an analytical capability to put
them into practice in a way that is:
• Informed by the needs of the service users
• Responsive to emerging questions
• Timely
• Robust
• Able to operate at scale
• Sustainable, providing an ongoing resource for the NHS
• Underpinned by a rigorous approach to information governance
• Able to draw on new analytical methods (and indeed, contribute to them)
11. The response
The Improvement Analytics Unit, a new partnership
between NHS England and the Health Foundation.
Our aim:
• By 2019, create a unit to provide rapid feedback on the
progress being made to improve care and efficiency as
part of national programmes in England
• Inform both local and national decision making
regarding the improvement of new care models
12. The Improvement Analytics Unit
Data
management
team
Statistical team
Leadership and governance
Intermediaries
with local and
national NHS
teams
performing
‘analytical
advisor’ role
Exchange with
universities to
draw on latest
methodological
thinking
Partnership between NHS England and Health Foundation
13. The first output
Principia is a multi-
speciality provider in
Rushcliffe
Nottinghamshire
From April 2014,
Principia introduced an
enhanced support
package for residents of
24 local care homes
14. The enhanced support package
• Alignment between general practices and care homes
• Advocacy and independent support from Age UK
Nottingham and Nottinghamshire
• Enhanced specification of general practice care for frail
older people living in care homes
• Improved peer-to-peer support from community nurses
for nurses employed within care homes
• A programme of work to engage and support care home
managers
15. Rates of hospital admissions for care home
residents per year
Note: The figure is based on the linked care home and hospital data for August 2014 to August 2016, and is for residents aged 65 or over in the six comparison areas, namely Harborough, Blaby,
Test Valley, South Cambridgeshire, Chelmsford and Brentwood. We excluded Rushcliffe since the aim of this figure is to illustrate baseline admission rates, and admissions in Rushcliffe might
19. Results
Principia
residents
(number per
person per
year)
Matched
residents
(number per
person per
year)
Relative
difference
(adjusted rate
ratio)
95%
confidence
interval
A&E
attendances
0.74 1.02 29% lower 11% to 43%
lower
Emergency
admissions
0.64 0.78 23% lower 3% to 39%
lower
Potentially
avoidable
admissions
0.22 0.30 28% lower 0% to 49%
lower
Elective
admissions
0.11 0.13 29% higher 36% lower to
163% higher
Outpatient
attendances
1.99 1.85 11% higher 12% lower to
40% higher
20. Interpretation
Older residents in Principia care homes experienced fewer
A&E attendances and emergency admissions than similar
residents in similar care homes elsewhere.
The matched comparison group had similar age, gender,
health conditions and prior hospital utilisation to the Principia
residents, but might have differed in unobserved ways.
Assuming the two groups were comparable, the most likely
explanation is higher quality care for residents of the Principia
care homes. This might be related to the enhanced support,
some other aspect of care delivery or local context.
As you may know, THF is an independent charity committed to improving health and healthcare for people in the UK
We fund frontline service improvement projects, development programmes and research, as well as undertaking in-house policy, economic and data analysis
Here today to talk about some work that we’re currently doing to understand and help guide the development of the NHS’ new care models programme
This morning you heard about NCMs and some of the other interventions that feature in STPs.
These are sophisticated interventions in a health system that is not just complicated, but a complex adaptive system where there is unlikely to be a simple ‘cause and effect’ relationship between interventions and outcomes.
Understanding the true impact of these interventions can be difficult: the lack of an obvious control group, for example, as well as identifying the role of intervention (i.e. people normally receive these interventions when they’re ill, and they often get better so is that the intervention or just regression to the mean?)
So, hugely important to ensure we can learn about what’s working and where we need to improve.
A number of techniques have been developed to assess the impact of this sort of intervention
One approach is to identify the people who have received an intervention and then match them with a control group who haven’t
This aims to find a group of people who mirror the intervention group as closely as possible in terms of demographic profile, health characteristics, etc
We can then compare what happens to each group and effectively isolate the impact of the intervention
Example from 2013
Birmingham OwnHealth, a telephone health coaching intervention aiming to reduce hospital admissions by providing people with phone based support to manage their conditions at home
Delivered to nearly 2,700 people with heart failure, CHD, COPD or diabetes, who all had history of hospital use
These are the number of emergency admissions per patient for each of the intervention and control groups, in the months leading up to the intervention group being given the intervention
There are some differences between the lines, but they are a relatively good fit
So what happens next?
Emergency admissions for the control group goes down – regression to the mean – but not for the people who receive the intervention
This intervention unfortunately didn’t deliver on its objectives on this occasion
Not clear why – need further qualitative work to understand whether it’s the intervention or the context in which it’s been applied, and ID how to develop to overcome these
Doesn’t mean that thinking behind it and whole concept of health coaching is flawed, but important to identify unintended consequences to inform how initiatives can be developed to be more effective
And there’s an obvious read across to the new care models programme…
Need to understand the impact of NCM on patient outcomes and cost at two levels
First is local: understanding what does and what doesn’t work to inform learning and guide the development of service delivery models
Second is national: same intelligence can be used to help develop service frameworks, identify and reinforce key enablers and remove barriers to change
The big challenge here is not developing the methods to understand what impact new care models are having, but having the analytical capability to use them in a way that is meaningful.
Feedback needs to be informed by the needs of the services users – otherwise may not be remotely useful.
E.g. if my Fitbit tells me I’m not being active enough, I know I need to move more. But it also tells me I’m not getting enough sleep, which is pretty useless without help in understanding why that is
Also need to be able to respond and adapt feedback to help answer emerging questions, as well as timely, regular and robust
Aim for NCMs to cover 50% of England by 2020, so need to build capacity for the NHS to be able to sustain and develop this analytical capability at scale
Just as the NCM programme has taken inspiration from the development of Accountable Care Organisations in the USA, we’ve taken inspiration from the approach taken by the Center for Medicare and Medicaid Services which aimed to use rapid cycle evaluation to provide frequent feedback to the people involved in delivering new care models locally while also evaluating the outcomes achieved by the different models being tested
IAU has been setup as a partnership between NHSE and THF
Joint leadership and governance of the Unit, with dedicated data management and statistical teams based at THF
Plus intermediaries to link NHSE with local teams to ensure analysis meets local needs
And work with universities to ensure we stay in touch with developments in analytical methodologies
Alignment – one general practice to one care home
Advocacy – e.g. to support residents and their families in changing to the aligned general practice
Enhanced care – e.g. a named GP visited the care home every 1-2 weeks to meet residents and proactively review medications and care plans
Improved peer support – e.g. training courses and signposting to existing specialist community services
Engage and support care home managers – care home managers network and regular meetings with the CCG
We measured hospital activity for as long as possible, following moving to the care home:
A&E attendances
Emergency admissions
A subset of ‘potentially avoidable’ emergency admissions
Elective admissions
Outpatient attendances
The percentage of nights spent as a hospital inpatient
Also, for the subset of residents who died, we assessed:
The percentage of deaths that occurred outside of hospital (included because the enhanced support included an element of end-of-life care planning)
Intervention group:
588 residents from 24 care homes, aged 65 years or over, who had moved into care home that cared specifically for frail older residents between August 2014 and July 2016
Not previously residents of care homes, but had experienced inpatient admission to hospital within previous two years before moving to the care home.
Control group:
Selected six areas of England with similar demographic and socioeconomic characteristics and emergency admission rates to Rushcliffe.
Selected a ‘matched group’ of 588 residents on the basis of similar resident and care home characteristics
588 intervention residents and 588 matched comparison residents, each group tracked for just over 200 days on average following moving to the care home.
Principia residents attended A&E 29% less often than the matched comparison group
And were admitted to hospital as an emergency 23% less frequently
Fewer avoidable admissions too, but confidence intervals mean we can’t be confident that was a meaningful difference
No evidence of an impact on the number of nights spent in hospital, or the number of elective admissions or outpatient attendances
And Principia residents were just as likely to die outside of hospital
So what to make of this?
Principia residents had fewer A&E attendances and emergency admissions than the comparison group
It’s possible that the two groups may have differed in ways we didn’t account for, but assuming the two groups were comparable the only explanation of the difference is that Principia residents received better quality of care
Where next? In process of scaling this up – continue testing the approach with another NMC Vanguard
Early lessons include importance of relationships with end users, and that ”unmet need” for analytics is very clear