commissioning for complexity: exploring the role of system ...€¦ · 1 introduction this appendix...
TRANSCRIPT
Commissioning for Complexity: exploring
the role of System Dynamics in social care
APPENDIX: Key Elements of the Model
1 Introduction This Appendix should be read in conjunction with the NIHR School for Social Care Research Scoping
Review:
Miller R and McKelvie D
Commissioning for Complexity: exploring the role of System Dynamics in social care, 2016
The Appendix provides some more detail about the model described in the Scoping Review, mostly
by showing some inputs to and outputs from a typical model run.
The model was built using iThink® software (the academic version of the software is known as
Stella®), one of a number of computer packages specifically designed for System Dynamics modelling
and simulation (using stocks, flows and feedback loops).
For a more detailed account of System Dynamics and how it might be used in Social Care, please
refer to:
McKelvie D Modelling social care complexity: the potential of System Dynamics
Methods Review No 14, NIHR School for Social Care Research, 20131
1 sscr.nihr.ac.uk/PDF/MR/MR14.pdf
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
2
2 Contents The Appendix includes:
• A list, and brief descriptions, of the main model inputs, including
o Size and dynamics of the population of people with learning disability
o Breakdown of the population of adults with learning disability by a combination of
need and service state
o Dynamics of how long people spend in different states / crisis stages, including how
an Intensive Support Service might change things
o A simple service specification for an Intensive Support Service
o Cost assumptions
• Screen shots of the model showing how it reports outputs, mostly in the form of ‘behaviour
over time’ graphs, numerical displays, and exportable tables, focusing on the model
‘dashboard’ with
o Changes in the population
o Detail of how adults are distributed across crisis / service states
o Utilisation of the Intensive Support Service Capacity
o Financial outputs, comparing cost of doing nothing with implementation of Intensive
Support
o Table Outputs
3 Model Structure For a full description of the model structure, please refer to the Scoping Review (and read it first).
4 Model Inputs Model inputs include:
• Basic population data
• Assumptions about how the population is distributed across need groups, and the rates of
movement between these groups
• Estimates of the cost of providing various services
• Estimates of the impact an Intensive Support Service might make on the use of, and
outcomes of, existing services
Any of the data inputs shown can be varied. Indeed, for the inputs where there is more likely to be
uncertainty, such as the impact of the ISS on crisis resolution or hospital length of stay / discharge
patterns, users of the model can perform multiple experiments along the lines of ‘what if this service
produced that outcome?
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
3
4.1 Basic Population Data
These inputs drive the behaviour of the main population ‘ageing chain’ representing all people with
a learning disability in the population. The first table describes the initial conditions of the model. As
the model moves forward in time, simulating the passage of 25 years, based on the ‘assumptions
about births and mortality’:
• More people with LD are born
• People move to the next age band
• Some in each age band die
In this version of the model, there is no migration into or out of the local authority (but this could
easily be added).
It would have been possible simply to use population projections as inputs to the model, but it is
generally better to simulate a dynamic process (of births, ageing and deaths) rather than use an
exogenous input.
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
4
4.2 Breakdown by Service State and Underpinning Dynamics
The focus of the model is the working-age adult population, aged 16 to 65. This represents one of
the population bands in the main ageing chain, and it is broken down into more detail, based on
‘service states’.
The ‘breakdown’ table sets the initial (start of model) assumptions about how many people are in
each state.
The ‘main rates of movement’ table sets assumptions about the rates at which people make
transitions:
• From the simple to the ‘with complications’ group
• The mean duration of the ‘with complications’ state (after which people return to the simple
state); note that although the mean is used as the input, the model actually simulates a wide
distribution of durations around that mean
• The rate at which people in the ‘with complications’ state experience a crisis, being an event
that might precipitate a hospital admission or placement in a care setting
(what happens next, in terms of how crises are resolved, is shown in the next table)
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
5
4.3 Movements through Crisis States (and impact of Intensive Support)
The main purpose of this model is to explore what would happen across the whole system if there
were to be an Intensive Support Service introduced, which would:
• Support people in crisis so that they are less likely to need hospitalisation or care
• Support people in hospital or in care so that they return to living in the community faster
Note that these inputs also set some baseline assumptions about what currently happens (without
an ISS).
As shown below, in this run of the model, we assume that the average duration of a crisis is 6
months and 20% of people in crisis make a transition to local hospital. But if Intensive Support were
available, only 10% would need admission, and the duration of a crisis would be only 2 months.
Similar assumptions are made about the duration of hospital admissions and probability of transition
to care, with and without Intensive Support. The point of the model is to explore ‘what might
happen if we could offer Intensive Support?’
The amount of Intensive Support, in terms of wte staff employed, is covered next.
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
6
4.4 Service Specification for Intensive Support
These inputs describe the broad parameters of an Intensive Support Service in relatively simple
terms.
• Number of staff employed (wte)
• Maximum caseload per person
• Average duration of service
o Varying by where the service user is
The model therefore allocates Intensive Support to people in the different crisis states, depending
on whether there is capacity available. The question then becomes one of what it might cost to
provide Intensive Support to people who might benefit, and what might be saved in the cost of
hospital and care, including out of borough care.
4.5 Costs
Cost assumptions are relatively broad brush rather than detailed. Intensive Support is based on a
wte cost including overheads, and hospitals and care placements are costed based on a rate per
Occupied Bed Day, as shown.
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
7
5 Running the Model and Viewing / Interpreting Outputs To operate the model, including entering the inputs listed above, there is a user-friendly interface.
The ‘dashboard’, shown below, is the home screen of the model, providing an overview of the whole
system.
This model dashboard is designed to be displayed on a computer monitor or, ideally, projected on a
large screen in a group model building workshop. It provides a graphical overview/summary of the
model’s overall behaviour in a specific ‘run’ of the model. To get more detailed output, including
figures, the user should click on the buttons at the top of the screen. Typically, the model is paused
every year, and users may choose to vary an input (for example, staff employed in the ISS) before
resuming the run.
Brief descriptions of the dashboard elements are as follows:
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
8
5.1 Population
Both graphs show a breakdown of the
population of people with LD by age
group.
The upper graph plots a time-series
graph of the four age bands over a 25
year period.
The histogram below shows a
breakdown of the population at each
point in time. The histogram updates
as the model runs. Because this
snapshot shows the model at the end
of a 25 year run, the histogram
summarises the population at that
end-point.
The numbers under the graph
translate the histogram into numbers;
again, the numbers displayed here are
the values at the end of the
simulation run displayed.
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
9
5.2 Detail about Adults with LD in each Crisis State
This shows the population of adults with LD broken down by the number using each service (or in
crisis state). The graph shows the change over time (clearly showing the decline in use of care out of
borough) and the numbers display the breakdown, updating as the model runs, so, in this snapshot,
they show the breakdown at the end of the model run. The ‘adults with ld’ and ‘adults with
complication (in community)’ numbers are not plotted on the graph, which includes only the service
states that are likely to be affected by the Intensive Support Service.
To enable comparison, the second set of numbers (labelled [c] for control group) show what the
numbers would have been if the Intensive Support Service were not deployed. These details are also
available on the graph; a model user can toggle between these two versions of the graph:-
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
10
This helpfully shows that even in the base case, without the ISS, there is expected to be a reduction
in use of out of borough placements, based on policies already being implemented.
Perhaps worth noting: the total number of adults with LD figure is higher at the end of 25 years if the
ISS is deployed, as in run[e]. This is because the numbers in crisis states are lower, and these states
are assumed to have a higher mortality rate. One of the impacts of the ISS has therefore been to
prolong life expectancy.
5.3 Utilisation of the Intensive Support Service against Capacity
This part of the dashboard shows how the ISS Capacity is used over time. In this run, the capacity
(line 2, maximum who can use the service at any point in time) is 18 (6 staff multiplied by max
caseload of 3). At the start of the model run, capacity is fully utilised and there is a waiting list
(people referred to service who are not able to access it). As time passes, the waiting list reduces,
and, after about five years there is spare capacity and hence no waiting list. In fact, the number of
staff could probably be reduced during the run, but this was not implemented in the simulation.
Once again, the numbers for the variables plotted are displayed below the graph. By the end of the
run, 0 people are waiting and there is a caseload of 11 against a maximum possible of 18.
Commissioning for Complexity: exploring the role of System Dynamics in social care:
APPENDIX: Key Elements of the Model
11
5.4 Finance
This graph plots annual expenditure, on the
items listed in section 4.5, comparing the
base (no ISS) run [c for control] against the
current scenario [e for experimental].
The numbers below the graph show annual
expenditure in £ millions. These numbers are
rounded to the nearest million here,
showing that, with the introduction of the
ISS annual whole system costs have reduced
from £47m to £43m by the end of the 25
year run.
The lower graph plots the difference
between these numbers. A positive value
means that the ISS produces a saving. In this
example, by the end of the run the whole
system cost is £4.484m per annum lower, as
shown in the number below the graph).
Because only one variable is plotted
(multiple times) on the lower graph, it is
possible to run the model repeatedly, using
different assumptions, and plot the ‘saving
pa’ each time, to facilitate sensitivity analysis
(not illustrated).
5.5 Table Outputs
With a dynamic model used in a workshop setting, most users prefer to view ‘behaviour over time’
graphs, rather than tables, but any of the outputs shown here, and more, can be shown in that way,
and these in turn can be exported to a spreadsheet, so that they can be used by someone who does
not have the model software installed.
© Douglas McKelvie, the Symmetric Partnership LLP, 2016