Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Care and Health Improvement programmewww.local.gov.uk/chip Care and Health Improvement programme
Predictive Modelling Round Table
Output Pack17th June, 2019
Care and Health Improvement programmewww.local.gov.uk/chip
Section 1: Overview
Care and Health Improvement programmewww.local.gov.uk/chip
Introduction
• On 17th June 2019, the LGA ran a workshop attended by 35 stakeholders
from across the country, who represented:
– Local Authority Commissioners
– Local Authority employees with specialist skills in a wide range of area including;
finance, data modelling and reporting and performance.
– Other local, regional and national stakeholders
• This workshop was facilitated by Doug Sheperdigian (Atlantic Customer
Solutions), who worked with LGA to design the structure and content
• This document, as with the workshop, is in 4 sections:
– Overview: Introduction and set up for the day
– Benefits: What would a predictive model help us achieve?
– Features: What are the components of a good predictive model?
– Next Steps: The pilot activity and final approach
Care and Health Improvement programmewww.local.gov.uk/chip
Guide to work through this document
• The outputs from the workshop have been set out in this document in a way
that can be read and understood by anyone, with the following caveats:
– Outputs will make most sense to those who attended the workshop and / or have
some understanding of this work. If in doubt, check with LGA colleagues
• There are three types of slide in this document:
– Presentation given at the workshop – Slides with red headings
– Guidance and instruction at the workshop – Slides with blue headings
– Information produced by at or before the workshop – Slides with green headings
– Slides that provide a narrative for this document – Slides with black headings
• In each section, you will find the outputs as produced on the day.
– Posters with added post-it notes, comments and “likes” in the form of hearts.
– Printed text on feedback posters is the output from interviews with key
stakeholders conducted in advance of the workshop
– Each section includes a brief summary of the key outputs
– Hearts are colour coded to indicate a perspective
Care and Health Improvement programmewww.local.gov.uk/chip
Workshop AgendaTime Topic Who Slide Pack Ref.
10.30 Registration and Coffee
10.45 Safety Briefing Brigid Day
10.50 Welcome Hazel Summers – Chair of Project Steering
Group
11.00 Ground rules Doug Sheperdigian – Facilitator
11.05 Who we are (Introductions) Facilitated session
11.15 Where are we now? Brigid Day and Rachel Carter Slides 8 – 11
11.25 Benefits – What should a good predictive
modelling approach help us achieve?
Working in small teams to gain the advantage of
all perspectives in the room
Slides 13 – 15
11.45 Report back Headlines from each team Slides 16 – 24
12.00 Features – what are the key components of
a good predictive modelling approach?
Working in small teams to gain the advantage of
all perspectives in the room
Slides 25 – 27
12.30 Report Back Headlines from each team Slides 28 – 38
12.45 Individual inputs Capture views
13.00 Working Lunch Activities will continue!
13.30 Intro to Next steps Leon Goddard Slides 39 – 42
13.40 Next steps and pilot input Working in small teams to gain the advantage of
all perspectives in the room
Slide 43
14.15 Report Back Headlines from each team Slides 44 – 53
14.45 Closing remarks and feedback Doug Sheperdigian Hazel Summers Slides 54 – 56
15.00 Finish
15.10 Market Analysis Philippa Lynch
15.30 Informal Networking (unstructured)
Care and Health Improvement programmewww.local.gov.uk/chip
Start to the workshop
• There was a warm welcome from Hazel Summers, who described what she
hoped to get out of today’s session
• Doug Sheperdigian, our facilitator, then explained how he would run this
session in order to get the most value from it.
• Doug also asked for permission to manage everyone’s time and the group
were happy to agree to this, some by nodding their head, others did so
telepathically.
• Brigid Day and Rachel Carter then gave an update on the project to date …
Care and Health Improvement programmewww.local.gov.uk/chip
Aims and benefits of overall project
To build on successful work progressed by some regions and councils to better
understand local care markets so councils can:
• Have access to comparable consistent market supply and demand
information within and across councils/regions
• Identify and manage better adult social care market risks
• Use tools to aid commissioners planning to meet demand and assist with
budget planning
• Improve commissioning plans and strategies
Care and Health Improvement programmewww.local.gov.uk/chip
Some national contexts
Increasing focus on commissioning ……
• CQC area system review findings – more planned
• Continuing major funding and market supply and sustainability challenges
• Green Paper ?????!
• NHS interest/joint working on social care commissioning
Care and Health Improvement programmewww.local.gov.uk/chip
Project Stages – proof of concept
1. Collection of dataset to create a market analysis tool on LG Inform Plus
enabling analysis and comparison within and between councils and
regions
2. Develop predictive modelling tools and support, to inform and enable
modelling of future markets at local, regional and national levels
3. Bespoke support to councils and regions to develop/refine their
commissioning plans & Market Position Statements
• 80 councils engaged
• Built on previous ADASS regional work
Co-designed data collection
82 councils submitted
Councils submit
data
Market Analysis Tools developed
Next steps
• Target June
• Show & Tell national findings to regions
First tools
available
• Are tools useful?
• Consider next steps
Review
• Forecast demand
• Understand options
• Decide markets needed?
Define aims
Models developed by councils and regions
Research
Workshop to plan coproduction approach
Options appraisal
report
Pilot phase across regions
Share progress
NEXT STEPS
Evaluation
Plan and agree future…
NCAS Conference
Support offers continue
Where we are now
Predictive
Modelling
Market
Analysis
Sept to Nov May to June July Nov- Jan
2018 → 2019 → 2020
Broad Predictive Model aims•Local focus - gold standard framework
for modelling with facility for local
‘deep dives’
•National focus - facilitating standard
tools to help councils & regions &
make national case for funding
Option 1
• No change
• No further action
Option 2
• Scale up one existing model
Option 3
• Combine best from council models
Option 4
• Develop a new model
Care and Health Improvement programmewww.local.gov.uk/chip
Predictive Modelling – 3 key areas
Section 2:
Benefits
Section 3:
Features
Section 4:
Next Steps
Slides:
13 – 24
Slides:
39 – 53
Slides:
25 – 38
Care and Health Improvement programmewww.local.gov.uk/chip
Section 2: Benefits
What should a good predictive
model help us to achieve?
Care and Health Improvement programmewww.local.gov.uk/chip
Benefits Roundtable
Jupiter: Regionally
Mercury: Locally Venus: Nationally
Pluto: Market Shaping
and TransformationSaturn: Finance
Neptune: Market
Sustainability
Care and Health Improvement programmewww.local.gov.uk/chip
What should a good predictive model
help us to achieve?
• Market Shaping and Transformation
– Achieving the right type, level and location of provision.
– Could seek gradual improvement or transformation
• Market Sustainability
– Matching provision with capacity and avoiding provider / market failure
• Finance
– Using financial modelling to make better projections and use of resources
• Locally
– Helping authorities, stakeholders and organisations in local areas
• Regionally
– Supporting cross boundary planning, commissioning and delivery
• Nationally
– Supporting the national policy agenda and financial support
Care and Health Improvement programmewww.local.gov.uk/chip
Benefits Report Back
Care and Health Improvement programmewww.local.gov.uk/chip
Pluto: Market Shaping
Care and Health Improvement programmewww.local.gov.uk/chip
Neptune: Market Sustainability
Care and Health Improvement programmewww.local.gov.uk/chip
Saturn: Finance
Care and Health Improvement programmewww.local.gov.uk/chip
Mercury: Locally
Care and Health Improvement programmewww.local.gov.uk/chip
Jupiter: Regionally
Care and Health Improvement programmewww.local.gov.uk/chip
Venus: Nationally
Care and Health Improvement programmewww.local.gov.uk/chip
Supports
opportunity for
regional
collaborative
commissioning
Useful for small
unitaries and
provide evidence
base for planning
at sub-regional
basis
Inform
regional SLI
programme
Future ability to
drill down into
granular to
identify best
practices/
service models/
outcomes
Accept the
differences for
different type of
care models e.g.
sti/ extra case may
require 10-15 years
Granularity
of data to
change
model/ trend
Provide key
intelligence to
inform STP/ICS
regional strategic
partnerships
Does this relate to
market oversight
and contingency
planning… Not
sure
Prompt
consistent
data
definitionsConsider
issues at the
boundaries of
the regions
e.g SW
London and
Surrey
Myth
busting
at a
local
level
Include sub-
regional as
markets are
often at that
level
Enables
“what if”
modelling
Encourage
s working
across LA
boundaries
Doesn’t need
to be 100%
correct, just
better than
now.
Inform
and
monitor
Regional gives
better
population
evidence and
see trends
Caution around
inappropriate
usage of
outputs
Do the
model/struct
ure once
not x no.
times
Cross boundary
approach
minimises cross
boundary impacts
where one LA
does something
affecting another
Sharing best
practice,
more easily
Support
working at
ICS/CCG
landscape
Core data
can be used
@
regional/sub
regional and
local level in
local
systems i.e.
P.A.M.M.S
Identify gaps
to meet short/
medium/long
term
Locally Regionally Nationally
This page contains ‘difficult to read post-its’ from the 3 previous slides
Locally, regionally and nationally – cont.
Care and Health Improvement programmewww.local.gov.uk/chip
Benefits: Summary & Initial Analysis
• The workshop teams identified
and prioritised 105 Benefits*
categorised by:
– Locally – 45
– Regionally – 41
– Nationally – 19
• A database has been prepared
as a separate output, capturing
all benefits.
*Further work can be done to cluster and de-duplicate these
Top 5 Local Benefits
No. Scope Comments
L3 Local Better commissioning
L1 Local Better market position statements
L5 Local Long-term signalling to market
L4 Local Better understanding of alternative
L15 Local Commissioners understand their role in changing demand
Top 5 National Benefits
N3 National Able to support requirement for more funding with Treasury
N1 National (Robust) Evidence based dialogue with DHSC
N12 National Potential to consider all ages in the analysis
N2 National A means to align ADASS and LGA in national dialogues
N4 National Potential to analyse Self funders
Top 5 Regional Benefits
R3 Regional Able to benchmark
R5 Regional Improve ability to deliver 10 year plans
R4 Regional Support regional scorecards
R32 Regional Understand current position + trends and movements
R7 Regional Prompt consistent data definitions
*Further work can be done to cluster
and de-duplicate these
Care and Health Improvement programmewww.local.gov.uk/chip
Section 3: Features
What are the components of a
good predictive modelling
approach?
Care and Health Improvement programmewww.local.gov.uk/chip
Benefits Roundtable
Jupiter:
Communications
Mercury: Infrastructure Venus: Outputs
Pluto: Information Saturn: Technical
Neptune: Resources
Care and Health Improvement programmewww.local.gov.uk/chip
What are the key components of a
good predictive modelling approach?
• Information
– What data, information and evidence must be included in the approach?
• Resources
– What type of resources (people, IT etc) are required and what are the costs?
• Technical
– What analysis, projections and programmes must support the approach?
• Infrastructure
– What needs to happen to ensure the right organisational structures, cultures
and approaches exist to make best use of this work?
• Communications
– What should we explain about this work, about how key stakeholders can be
involved in the development and about the approach and how it works?
• Outputs
– What must the model predict? What are the outputs
Care and Health Improvement programmewww.local.gov.uk/chip
Features Report Back
Care and Health Improvement programmewww.local.gov.uk/chip
Pluto: Information
What do first people then cost
First people data
People at front door + self funders
Standardsnational
Key question: how many people in system- Now- 2-3 yrs
Target audience, which customer
Risk safeguarding data
What is starting point? Whole population or social care population
Housing Data
Stress test End of
life
Length of stay reducing
HES data linked to death registration data
Self funder info helps care home risk
Inputs Technical Aspects Outputs
This page represents the post-its from the
previous page – transcribed for legibility
Pluto: Information cont.
Care and Health Improvement programmewww.local.gov.uk/chip
Neptune: Resources
Governance e.g. across housing and ASC
Good leadership and conceptual maturity across all participansgroups
Wider applications and use of data beyond ASC
Issue of integrating data.
Actuals vs predictionse.g. impact in 5yrs will dictate monitoring
Has to be understandable approach
Frequency depends on user group/ service
What it does:Step to inform planning – not a panacea
Sophisticated vs solid base for a model
Web based tool vs excel
Extraction of data
Repeat users
Age of entry
Average time in provision
3 years*
Housing data
Performance data support- Local- national
Provider input
Children’s and transitions data
Example of information about informal carers.- What info needed which can influence
ICT- Local- national
Skill mix of staff team –“no one is an expert in everything”
Inputs Technical Aspects Outputs
This page represents the post-its from the
previous page – transcribed for legibility
Neptune: Resources cont.
Care and Health Improvement programmewww.local.gov.uk/chip
Ability to bring
existing data
(from various
sources) into the
model, with an
agreed core
dataset.
Unique identifier
e.g. core
location I.D.
In all registered
managers to
keep up to date
Regular
(monthly)
data inputs
locally
Automated
collection
and delivery
with the
system
What if
planning.
“What if
scenarios”
Saturn: Technical
Care and Health Improvement programmewww.local.gov.uk/chip
Self
funders
running out
of assets
Commercialisation
– Sell information
Mercury: Infrastructure
Care and Health Improvement programmewww.local.gov.uk/chip
Jupiter: Communications
Care and Health Improvement programmewww.local.gov.uk/chip
Elephant
Clarity of definitions of
data.
Keep it simple and
linked to recognised
categories
Who are the
audience for this –
internally –
leadership team
and members
Co produce with
providers + public
solutions/
transformations to
alter trajectory
Risk of providers
exiting the market
permanentlyHow to get national
implementation
- ongoing comms +
engagement
Share resource
available to
providers to
consider markets
too – open data
HOW:
Markets <-> predictive
Data <-> modelling
Yes but is the
proposed
model going
to do this?
Key, simple info about
expenditure to help
communicate with the
public.
Inputs Technical Outputs
Need to establish if this: -is only a forecasting tool or if it is also real time monitoring -incorporate provider info/links to care market data.
Jupiter: Communications – cont.
This page contains ‘difficult to read post-its’ from the Jupiter slide
Care and Health Improvement programmewww.local.gov.uk/chip
Need to be able to
drill down to
understand “actual”
vs “forecast”
Cost
- Volume
- Complexity
- Market
prices
Workforce
- Volume
- Skill mix
- Supply
mix
Venus: Outputs
Care and Health Improvement programmewww.local.gov.uk/chip
Features: Summary & Initial Analysis
• The workshop teams identified
and prioritised 168 Benefits*
categorised by:
– Inputs – 77
– Technical – 54
– Outputs – 37
• A database has been prepared
as a separate output, capturing
all benefits.
Top 5 Inputs
Top 5 Outputs and Key Features
Top 5 Technical Aspects
*Further work can be done to cluster
and de-duplicate these
No. Scope Comments
I1 Inputs
POPPI/PANSI historical volumes and costs (2 years)
deprivation
I2 Inputs Workforce avaliability
I4 Inputs Inflation estimate
I5 Inputs Integrated with Care IT systems (all of them)
I3 Inputs Land costs
T1 Technical
Transparent mathematics (so that we can explain) - "must not
be a dark art"
T2 Technical Local judgement applied
T3 Technical Statistical techniques (e.g. logistical regression)
T5 Technical Core set of data -consistent
T5 Technical AI/machine learning
O1 Outputs
Future net change in demand by service area (e.g. OP - res care) - new service users -leaving the
service -total service users -level of need (complexity) -"going beyond simple demographics"
O2 OutputsMonitoring actuals versus prediction on a [real time, monthly] basis
O3 Outputs 5 year projection of demand and costs
O4 Outputs Cost (budget prediction) by service area "going beyond simple inflation"
O5 Outputs Communicates to laypeople and non - analysts
Care and Health Improvement programmewww.local.gov.uk/chip
Section 4: Next Steps
Pilot Activity and Final Approach
Care and Health Improvement programmewww.local.gov.uk/chip
Next Steps – Pilot Activity
Build on current knowledge and information
Seek to support decisions and design around final approach
1. Purpose and objectives
– What are we trying to understand and learn from the pilot activity
2. Design and development
– What type of activity will be undertaken and what areas will it focus on?
3. When and Who
– When should the pilot happen? Who should be involved? How authorities will be
selected? What approaches should be piloted?
Care and Health Improvement programmewww.local.gov.uk/chip
Next Steps – Final Approach
3 broad options that sit along a continuum
All options will be co-produced and based on learning and best practice
1. Single Model
– Complete and comprehensive model
– Expectation is that all authorities will use this model
2. Agreed Principles
– Key principles agreed by stakeholders
– Authorities to design their own approach, that follows these principles
3. Toolkit
– Include documents, specifications, guidance, hints and tips
– Authorities free to use what they want to deliver the approach they want
Care and Health Improvement programmewww.local.gov.uk/chip
Individual Exercise – What point do you
see on the continuum?
What is the Best approach?
What is most practical and achievable?
What will Actually Happen?
One Model ToolkitAgreed Principles
Care and Health Improvement programmewww.local.gov.uk/chip
Each Team has a Specific Assignment
Jupiter:
Option 1 –
Single Model
Mercury:
Option 2 – Agreed Principles
Final Approach – Please describe what you think
are the pros and cons of the different final
approaches – please state your preferred approach
Pilot – Please describe what you think any pilot
activity should look like; Use a flipchart to capture
your team’s views.
Venus:
Option 3 – Toolkit
Pluto: Purpose
and objectives
Neptune:
Design and
development
Saturn: When
and Who
Care and Health Improvement programmewww.local.gov.uk/chip
Next Steps Report Back
Care and Health Improvement programmewww.local.gov.uk/chip
Pluto: Purpose and objectives
Care and Health Improvement programmewww.local.gov.uk/chip
Pluto: Purpose and objectives cont.
Care and Health Improvement programmewww.local.gov.uk/chip
Neptune: Design and development
Care and Health Improvement programmewww.local.gov.uk/chip
Saturn: When and who
Care and Health Improvement programmewww.local.gov.uk/chip
Maximum benefit- National- Regional
Matching a national/ Regional NHS
Balancing whole population analysis vs those who use council funded services
Jupiter: Option 1 One Model
Care and Health Improvement programmewww.local.gov.uk/chip
“One Model” Pros and Cons
Pros• Balancing whole population analysis vs
those who use council funded services
• Maximum benefit
- National
- Regional
• Matching a national/ Regional NHS
• Build on best practice
• Economies of scale (limited resources)
• Consistency at a national level
• Single narrative
• National benefits realised
• The only one that works
• Benchmark (value of)
• National evidence to influence policy development
Cons
• One size may not fit all
• but it does really!
• It might hurt a bit
• Input data is restricted to what is
available across all LAs
Care and Health Improvement programmewww.local.gov.uk/chip
Mercury: Option 2 Agreed Principles
Care and Health Improvement programmewww.local.gov.uk/chip
Venus: Option 3 Toolkit
Care and Health Improvement programmewww.local.gov.uk/chip
Red=Regional
Yellow=National
Green= Technical, Finance
Blue=Commissioners
We asked individuals
to place a “like” in
answer to 3 questions:
• The majority chose the
“one model” approach.
• And, it was seen as
highly practical…
• However, the most likely
outcome attracted a wide
range of opinion.
Care and Health Improvement programmewww.local.gov.uk/chip
Final Discussion and Next
Steps
Care and Health Improvement programmewww.local.gov.uk/chip
We finished with a plenary discussion
• “Has opened discussion on scope”
• “Opened discussion for consensus and agreed way
forward”
• “More sizeable than initial thought”
• ”Narrow down focus and get focus agreed”
• “Last time real tension between local and national
purposes. Now it seems more aligned”
• “People want simple answer to complex questions”
• “Toolkit idea was “ok” but need best of both bits”
• “Need a consistent data set, clear on what we’re using it
for”
• “Not as simple as “one model”
• “Everyone’s at different points”
• “We’ve not separated starting points and assumptions,
weve looked at assumptions today but what’s the
starting point.?
• “We need to define what’s generating that percentage”
“What are the “process” parts of the model – what data
we want, what data are we putting it in, what are we
multiplying it 2by”
• The further up the food chain in the system, the more
inaccurate the data”
• “The starting point matters”
• “Potentially choose something that already exists rather
than trying to reinvent the wheel”
• “Use this as the starting point – test it – try it – analyse
it”
• “Consider what Alder have already done.”
• “Only at start of the process but thinking about
integration and health” What does this actually mean as
a bottom line to hospitals for beds etc…”
• “Considerations of different pressure scenarios included
in the model would help with this application link and
bridging that gap”
• “When we visited councils and regions we saw good
practices. Model is a puzzle, with different pieces, but
from different councils. Others had better bits than
others – call to effect etc. Principles more than
practicals – we already have enough to work with – don’t
need to start from scratch.”
• “What’s acceptable in terms of a model? – don’t want to
have to be polishing it forever”
• “Accurate usable meaningful working robust ease of use
– to develop credibility going forward”
• “Pilot has potential to fail quickly and allow us to move
on – pilot authorities may not be total representation of
all the potential participants”
• “Never going to be completely accurate but would be
happy with 80% good?? Way better than now!”
• “Find the right scale and ambition from the beginning
and not try to do too much too quickly”
• “Don’t run before you can walk but still aim high”
Care and Health Improvement programmewww.local.gov.uk/chip Care and Health Improvement programmewww.local.gov.uk/chip
Hazel brought us to a close, indicating
some proposed next steps
Pilots
• Identify at least one council per ADASS region to serve as a pilot
• Design and fund the pilot programme
• Deliver pilots (test different scope and complexity)
Communication
• Report back to working group from the Workshop
• Deliver a workshop at NCASC (November)
Governance
• Refresh the reference group and include predictive modelling as a key term
of reference.
Care and Health Improvement programmewww.local.gov.uk/chip Care and Health Improvement programme
Key Contacts:
Leon Goddard
Senior Adviser CHIP/LGA
Commissioning and Markets
07557 214985
Rachel Carter
Adviser CHIP/LGA
Commissioning and Markets
07827 083 942
For more about CHIP work on commissioning see
Commissioning and market shaping | Local
Government Association