The use of SD methodology to develop services for the
assessment and treatment of high risk serious offenders in
England and Wales.
Douglas McKelvie, Associate, Symmetric SD Ltd., The Grain Store, 127 Gloucester
Road, BRIGHTON, UK, BN1 4AF, +44 7939 634039,
douglas.mckelvie @ symmetricsd.co.uk
Savas Hadjipavlou, DSPD and TC Programme Director, National Offender
Management Service, Ministry of Justice, 2 Marsham Street, LONDON, UK, SW1P.
ADF, +44 20 7035 6922, savas.hadjipavlou @ justice.gsi.gov.uk
David Monk, Director, Symmetric SD Ltd., The Grain Store, 127 Gloucester Road,
BRIGHTON, UK, BN1 4AF, +44 1273 811092, david.monk@symmetricsd.co.uk
Samantha Foster, Senior Research Officer, DSPD and TC Programme, National
Offender Management Service, Ministry of Justice, 2 Marsham Street, LONDON, UK,
SW IP 4DF, 020 7035 6904, samantha. foster @justice.gsi.gov.uk
Eric Wolstenholme, Director, Symmetric SD Ltd., The Grain Store, 127 Gloucester
Road, BRIGHTON, UK, BN1 4AF, +44 7850 402864,
eric.wolstenholme @ symmetricsd.co.uk
David Todd, Symmetric SD Ltd., The Grain Store, 127 Gloucester Road, BRIGHTON,
UK, BN1 4AF, +44 7921 165510, david.todd@symmetricsd.co.uk
Abstract
In England and Wales, national government has introduced a range of policies for
dealing with the most dangerous offenders. These include new sentencing
arrangements, new treatment programmes, and enhanced supervision of those released
on licence. Policy makers needed to estimate the impact of this policy on the prison
population and to consider how much treatment and community supervision capacity
would be needed over time. They worked with a small SD consultancy to develop and
build a model using ithink software, to enable a variety of scenarios to be tested,
adopting a group model building approach. A number of staff received training in SD,
including model building. The paper outlines the policy background, model structure
and examples of scenarios. As well as being a practical application of SD to a sensitive
area of public policy, the project is an example of what can be achieved within a
relatively short intervention.
Introduction
System Dynamics is being used extensively by the authors to assist decision making and
integration of policy implementation along long service user flows crossing multiple
agency boundaries. The work has involved national level studies to influence
government policy on delayed hospital discharges (Wolstenholme et al, 2004a) and
more recently to assist local heath and social care communities in the UK to interpret
and apply national policy frameworks for older people (Wolstenholme et al, 2004 b and
c) and for mental health reform (Wolstenholme et al 2006). The system described in this
paper is at the interface between health care and criminal justice, the treatment of
dangerous offenders having severe personality disorder.
Background policy and motivation
In England and Wales, an aim of criminal justice agencies is to protect the public from
serious violent or sexual offenders. The impact on victims and their families is clearly
very significant. More generally these offences also have a disproportionate effect (to
their overall number) on public and political perceptions of crime, risk and fear of
crime. For our society they continue to attract high levels of media interest and public
protection is high on the political agenda.
The problem for criminal justice agencies is that these offenders are among the most
difficult to manage and perhaps also the least likely to rehabilitate. A significant
proportion, around 2,000 to 2,500 offenders in high secure prisons, can be characterised
as “dangerous and suffering from severe personality disorders”. They pose challenges
both in the criminal justice system and health services. Their assessment and treatment
remain contentious issues with uncertainty about the taxonomy of diagnosis,
assessment, treatment modalities and their effectiveness at reducing risk of further
offending and easing institutional management. Any attempt to provide services for this
group faces challenges of organisation and service delivery, with complex
multidisciplinary approaches being needed as well effective interagency working.
In the UK we started to tackle some of these issues through the Dangerous and Severe
Personality Disorder (DSPD) Programme (British Journal of Psychiatry (2007)
Supplement 49 — in press, full reference awaited). This is a range of pilot assessment
and treatment services hosted either in prison or secure hospital settings, with the stated
objectives:
To enhance protection of the public and improve mental health outcomes by
understanding better:
How to identify, assess and treat those who are dangerous and severely
personality disordered
The nature and challenges of treatments and service delivery involving multi-
disciplinary teams working across agencies
The extent to which treatment might reduce (or manage better) the risks of re-
offending and how best to move on those offenders who have benefited from the
programme, as well as those who have not
To strengthen the clinical, service delivery and policy evidence base in this area,
informing the options for future services, and the costs and_ benefits
Key Programme deliverables include four high secure pilot projects providing 300+
DSPD places (HMP Whitemoor, HMP Frankland, Rampton High Secure Hospital and
Broadmoor High Secure Hospital), and 75 medium secure and community places, as
well as a pilot for women offenders with severe personality disorders.
These various projects are intended to test out service delivery models and the possible
working of pathways for those that benefit from the interventions and those that do not.
In this context, therefore, the question about the shape of future services raises
significant issues about their organisation, particularly given that they are expected to
“sit” within the broader framework of prison and secure hospital provision. The high
cost of these services also makes it important that their planning is evidence based and
informed by realistic - and credible - projections of demand and effectiveness. The
complexities involved also mean that there is a need to communicate a clear and
coherent picture to various stakeholders. Given this range of challenges the SD
approach provides an effective framework for organising our experience and
information, and provides the ability to make projections while quantifying some of the
uncertainties involved.
A key issue for these services is the comparatively long periods that need to be factored
into the planning and our assessment of interventions. Sentences are either determinate,
or indeterminate — with release dependent on risk assessment, once a minimum “tariff”
has been served — generally in excess of 4 years. For this group of offenders current
treatments, too, are generally expected to be between 3 -5 years. This means that the
time frame for population projections needs to be long enough to capture the full “life
cycle” of people entering the criminal justice system and that the benefits could only
sensibly be assessed over a period of 10-20 years. Again the effect of feed backs and the
stock and flow organisation of models in SD lends itself naturally to this type of
problem.
Assessing the benefits also poses serious challenges. Our experience of these services is
limited. Yet we still need to think about the problem of how best to manage an
increasing long term prison population, what interventions might prove effective and
what impact we might expect in terms of public protection. The ability to test out
assumptions about the impact of different approaches, including of legislation is clearly
important.
For policy makers the ability to approach these complex problems in a systematic,
organised fashion, one that is capable of bringing together, as a whole, a view of the
entire system when most people who work in it glimpse only their part, is clearly very
attractive. However this on its own would be limited. The ability to quantify the flows,
the work involved, its impact on resources, and to compare between variants or
alternative approaches makes the SD approach compelling. Crucially, however, all this
depends on being convinced about the reliability of projections to build up our
confidence that the models accurately capture the essential elements of the processes we
are trying to describe and create credibility with other stakeholders in the organisation.
In this respect we were fortunate in having a long run of data on life sentence prisoners
(a significant subgroup) with which to validate the model.
Development and use of the model
Since DSPD is a pilot programme, a key issue is to consider what progress is being
made: are the programme objectives being met, and how should we mould the future
shape of services, particularly in the context of the broader strategic development for
offender services, as well as developments in the National Health Service? Our
approach here has been to draw together the available programme experience, in terms
of delivery and in the emergent finding of research or other studies, and to use the
model to explore some alternative scenarios, estimating the impact of changes in policy
of the introduction of new sentence types, treatment regimes and community
supervision arrangements on the population of dangerous offenders over time. The main
parts of the system across which dangerous offenders are distributed include prison,
high secure hospital, and community supervision (which may include people living in
the community and those living in more specialist resources within communities, such
as medium secure hospital or hostel). By representing this population in a model, policy
makers would have a tool that would enable them to test the impact of a variety of
scenarios on future numbers.
Given that an important element of the new policy was the introduction of new
treatment regimes within both prison and high secure hospital, the model would also be
used to support capacity planning for these treatment centres. Key constraints here
include recruitment of, and time taken to train, specialist staff to run the treatment
centres, as well as the pace at which the physical capacity of treatment services might
grow. The model would therefore need to simulate assessment and treatment capacity
growing over a period of years; it could not be generated simply by directing large
numbers of staff around the system.
Some of the main questions to be asked of a model would include:-
o how many prisoners of this kind can we expect there to be?
o how will they be distributed across the different stages of sentence?
o what programme capacity would be required under various scenarios including
constraints on growth?
o. if the programme successfully treats people, what impact will that have on the
rate of release of these prisoners, and how many will require supervision in the
community?
Methodology
The project took place over a six month period starting in April 2006 and had the
benefit of experience of a system dynamics approach being used in 2005 (unpublished)
in relation to the same population of offenders, but at a regional level. The modelling
team comprised a model developer / facilitator and a facilitator/project coordinator,
whose work was overseen at all times by an expert adviser. The main roles of each
were:-
Model developer / facilitator
To attend all meetings, conduct additional fieldwork, build different iterations of the
system dynamics model, produce model documentation, report back to the main
stakeholder groups, design and deliver training to members of the client team.
Project coordinator / facilitator
To agree terms of reference with the client, attend and jointly facilitate main
workshop sessions, and ensure delivery of the main project outcomes
Expert adviser
To oversee and quality assure the model development, attend main stakeholder
meetings, lead on designing and delivering training to members of the client team.
The modelling process can best be characterised as a brief, group model building
approach, with flexible involvement of different members of the group. This collegiate
approach reflected the different roles and remits of the various project stakeholders.
There were number of groupings, each with overlapping membership:-
The policy team
This consisted of the full-time civil servants whose job it is to develop and oversee
national policy for dealing with DSPD offenders. The team comprises a mixture of
policy makers, practitioners (such as psychologists) and researchers. Meetings were
held sometimes with the whole group, but more usually either with the team
managers or the practitioner / researchers.
The role of the team managers was to influence the structure of the emergent model.
The most senior manager, having a grounding in operational research, became
increasingly involved in detailed technical aspects of model-development, especially
in establishing the initial conditions for the model, and ensuring its accuracy in
replicating the historical numbers of prisoners in the system.
The practitioner / researcher sub-group also influenced the structure and design of
the model. They contributed most of the detailed data inputs. As the project
developed, they were identified as the main personnel to be trained in using the
model. This training would include an introduction to system dynamics, specific
training around the use of Ithink software, and a detailed introduction to the model
itself.
Stakeholder Sub-groups
Stakeholder sub-groups were ad hoc groupings of those responsible for managing
particular parts of the system, typically either prison or secure mental hospital
managers. Their main role was to advise on the stock-flow representations of the
parts of the model dealing with the structured programme element of a prison
sentence, or hospital stay.
The programme board
The board has formal responsibility for the policy, and includes many of the
groupings already identified, whether as full members or as staff in attendance at
meetings. The board received updates on the development of the model, and full
presentations of the model as it emerged.
Wider stakeholder meeting
At the close of the project, the model was presented to a wider stakeholder group,
comprising many of those who had been involved in the project along with a wider
range of practitioners and researchers, including those responsible for release of
offenders and their subsequent management in the community.
This configuration of groups differed somewhat from that encountered in other
modelling projects within the public sector. The precise form taken by a group-
modelling project (Vennix, 1996) will vary according to the preferences of a modelling
team and its previous experiences of conducting similar projects, but also with the
nature of the client organisation, its established formal and informal structures, and its
political context. Sometimes it is possible for a modelling team to propose and
implement a very structured methodology (Luna-Reyes et al, 2006) with prescribed
roles for a modelling team and expectations of a client.
What does seem to be the key to the success of any group model-building project is the
consistent involvement of the same small (around 5-10 members) group throughout
each stage of a project. Here, that was achieved through the commitment of the staff
group, who were involved in the conception, development, first iteration, revision, and
completion of the final version, of this model. Consistency was achieved, even although
the project involved a sequence of meetings with various ad hoc groupings.
The Model
There were two main stages in this project, each lasting approximately three months.
The mid-point was marked by a presentation to the project board, the result of which
was that various additions and modifications were proposed. Whereas the first iteration
of the model was largely built by the modeller/facilitator the second benefited from a
more hands-on role being taken by the staff-team.
Two models were built, using Ithink software. Only the second is described in detail.
The client required a model with a long time horizon; prison terms and expected
treatment periods for these offenders would be lengthy. The initial conditions in the
model should be as accurate as possible. The preferred means of achieving that was to
build up the prison population by running the model to the start point of the new policy
(2003) over a forty year period. Although it would have been possible to choose a start
time for the model of 2003, and input initial values which had been calculated
separately, the staff team’s preference was for the model itself to generate the 2003
values. This had the potential of creating a more flexible backdrop against which to test
the impact of the treatment programme, its timing, the relative contributions of
assessment and treatment to benefits measured by serious offences prevented or access
to services. So the model runs for a 60 year period, comprising the 40 years prior to
2003 (during which the specialist treatment programmes did not exist) and the 20 years
following implementation. The graphs shown only cover the last 20 years.
The model represents the male prison population in England and Wales of the sentence-
types most likely to include men diagnosed as DSPD. These include the new
indeterminate sentences (IPP) and those sentenced to life imprisonment (a particular
type of indeterminate sentence). The model also represents a subset (DSPD males) of
those on determinate sentences (those sentenced to 4 years or more), but not
everybody on these sentences, in which case the model would include almost the whole
prison population. The main difference between indeterminate and determinate is that
prisoners sentenced to the former can only be released if the Parole Board allows this.
Men on determinate sentences must be released after serving their time. The term
“tariff’ means the minimum time that must be served before being eligible for parole.
The model uses arrays to represent these three sentence types.
Figure 1: Simplified Version of DSPD Stock / Flow Structure
stanetorod to cn programme
post programme
secur hospital in hostel inhospital
oso pog
assessed f
fect stage of tweatrert vitor complting tari post tet
seriance pegrarme poyamme on programme post programme peaeuey
ou LI {+2 H+ +} 2
sentenced to aseerement sutabletr 7! pleced camgleted mica enn
cage programme ost p
compltirg ttt post at
deco nenpiogemre non programms
tat places pas 1
LI ending tariff Li released non prog
reassessed as being dapd p
snot suteble for programme ending tart
er ‘ron dspd teleased non dspd
} =a)
sereened out a9 non dps comdltng tit posta
ron dsad
ron deed
This illustrates the main movements of men from being sentenced, screened out as not
DSPD (some of whom are later reassessed as being DSPD and referred back for
treatment), assessed for treatment, through treatment programme in prison or hospital,
and then to completion of sentence tariff. On completion of tariff, prisoners are eligible
to apply for parole, but this is subject to a rigorous risk-assessment. Most will spend
many years in the post-tariff stage, and some might never be released. Following
release, they are supervised in the community.
Figure 1 shows that there are effectively four routes through the system, resulting in
four different “post-tariff” stocks:-
o Those defined as DSPD who completed a treatment programme in prison (in the
main model, not shown here, further subdivided between those who successfully
completed the programme and those who did not)
Those defined as DSPD who completed a treatment programme in secure mental
hospital (if they meet the terms of the Mental Health Act), further subdivided as
above
Those defined as DSPD who, for various reasons, either a lack of programme
capacity or because the individuals had not been considered suitable, did not go
on a treatment programme
o Those who are non-DSPD, so obviously do not go on the treatment programme
Main Feedback Loops
The main feedback loops operating within this system are shown in the CLD below
Figure 2 — Causal Loop Diagram of DSPD System
newly sentenced
s
recidvisis OS.
5 dangerous
s prisoners
number at liberty AR} s ee
onlicence er finish tariff
:
a0
. number of
wed Jos
‘B) a stillimprisoned fraction accepted
cena
inate
reoffending rate low risk °
: 6
aman ica pn
low secure status. i
lasting impact 2
arremyne
fraction of sentence programme.
s 5 tpenton, programme -——— “ jength
programme impact ‘a
tariff (minimum time
to be served)
qualified Staff pet ay
The main feedback loops concern:
1. The capacity constraints of treatment programmes (availability of suitably
specialist qualified staff, availability of suitable facilities, how many can be in
treatment, timing of treatment towards beginning / middle/ end of sentence,
fraction of sentence spent in treatment)
2. The impact of treatment on levels of risk (if treatment is effective, men will
progress towards lower security categories faster, and become eligible for
release sooner) — in the longer term, fewer of those released on licence will re-
offend
Client representatives probably felt more confident about representing the treatment
capacity constraints rather than the treatment impact effects. This is a highly sensitive
area, and the time-scales over which the new policy would have an impact amount to
many years. At such an early stage in the implementation cycle, there would be no data
about the long-term impact of the programme, and it would be understandable for
policy-makers to be cautious about estimating this. Nevertheless, one of the key
dynamics of the factors governing release from prison is the level of risk (for which a
proxy measure would be the distribution of prisoners across formal “security
categories”).
A More Detailed Description of the Model
The detailed stock-flow structure of the Ithink model is represented in Figure 3. It is not
necessary for readers to be able to make out the individual variable names which will be
explained in a series of smaller diagrams describing each stage in the process.
Figure 3: Detailed Stock — Flow Structure with Main Stages Superimposed
Mental Health Act Route
Programme
Screening Assessment Pre Post Community
Tariff Tariff Supervision
"Achievement of a low-security classification is a necessary — but not a sufficient — condition for release
on parole
2 J h=
Figure 3.1 Screening Stage
First Stage
Screening
rg >
Newly to Wait for
Sentenced Assessment
ra
to Non DSPD Route
Men enter the system at the point of being sentenced. The newly sentenced rate is
exogenous, based on historical data and assumptions about how courts will use the new
indeterminate sentences. During the first stage of the sentence (which might actually
take several years for those on life sentences), men who do not meet the criteria for
DSPD (the majority of prisoners) are screened out and complete their sentence without
going on the treatment programme. The remainder proceed to be assessed for the
programme. Those going on to the assessment stage include those who are considered
as possibly meeting the criteria for DSPD.
2] [=
Figure 3.2 — Assessment Stage
to Wait for
Hospital
to Wait for
Programme in
Prison
Waiting for Being
Assessment Assessed
to Walt for Starting
Assessment Assessment
Assessed as Hot
For Programme
Non DSPD in
Prison
Ly]
Lt
To be Reassessed
for Programme
At the assessment stage, men who meet the DSPD definition and who are deemed
suitable candidates for the treatment programme proceed to the stage of waiting for a
programme place. Assessment is capacity constrained. Men might therefore have to
wait to be assessed. Under extreme conditions (hopefully not replicated in reality), some
men would wait so long that they missed out on the treatment programme, their
sentences having been spent waiting for assessment, represented in Figure 3.2 as the
flow “assess to stepaside” . The rule applied in the model was that men whose tariff
ended whilst they were still waiting for assessment would proceed to the post-tariff
stage without having been on the programme phase.
mead Wy
Figure 3.3 — The DSPD Treatment Programme — Prison Version
Wait for
Prison On Prison
Programme Programme
ie | 1
to Wait for starting Prison to Stepdown
Programme in Programme
Prison
169)
to Stepaside
39)
To Stepaside owing
to lack of Programme
Capacity
Some men go from assessment to “wait for treatment in prison’. In reality, there are
variations on this dynamic. Some treatment programmes admit men in order to assess
them, meaning that if the outcome of assessment is a recommendation “to treat” the
prisoner is already occupying a treatment place, and so proceeds to the full treatment
programme with no further delay.
==
The structure represented in the main model is that those who are assessed as requiring
treatment in a prison setting will then have to wait for a treatment place. Treatment
places are capacity constrained, and throughput is governed by the number of places and
the average length of stay in treatment (5 years, although this would be a variable to be
adjusted reflecting differing approaches to treatment). Onward movement from the
treatment programme is either to “stepdown”, meaning that the programme has been
effective and men will move down through security classifications at a faster rate than
the untreated population, or to “stepaside”, meaning that treatment has not been
effective and men will move down through security classifications at a slower rate or,
perhaps not at all, given that the risk of re-offending had not changed.
Note that if treatment places are insufficient to meet demand, and waiting times for
treatment get so long that men would still be waiting for treatment even when their tariff
is complete, men are taken off the waiting list into the post-tariff state, and with a
“stepaside” profile, because they were untreated. This is similar to the dynamic
described under Assessment. This route provides a safety valve in the model under
extreme conditions, and would not be expected to be replicated in a programme of
treatment services designed to meet the anticipated demand. However if the treatment
programme is not extended beyond the limited pilot stage this is a strong possibility.
The model allows for various inputs to be made setting treatment programme capacity.
Because availability of trained staff is limited, the treatment capacity is normally
modelled as having an initial amount, with more phased in over a ten-year period, in
accordance with policy makers’ expectations about resources.
-14-
Figure 3.4 — The DSPD Treatment Programme — Hospital Version
to long discharge
stay DO on supervision Oo ending CPA
in Hopital Commurity
Long Stay supervision CPA,
awaiting
Hospital on Hospital
srogramme Programme
e L Lh 0
toWat ‘Starting return to Prison.
for Hospital Hospital a5 St=pdown
Programme
7]
return to Prison
a5 Stepaside
Some who require treatment are transferred to a treatment programme in high secure
hospital. This is only possible if they meet the criteria for being held under the Mental
Health Act — under English legislation, having a personality disorder may not be
sufficient grounds for detention, and clinicians also have to be satisfied that the
individual is “treatable”. This pathway has almost identical characteristics to the prison
programme route described above. At the end of the programme, men either return to
prison (“stepdown” for those whose treatment has been effective, “stepaside” if not).
Others might never return to prison but remain in hospital (but not on the special
programme) as long stay patients, eventually being released to community supervision
under the “care programme approach”.
It is probable that transfers to hospital will be more readily sought for prisoners on
determinate sentences — the hospital route provides a means whereby the most
dangerous can be held without limit of time, but only if they have a mental disorder of
sufficient severity.
see Ni ae
Figure 3.5 Pre-Tariff, Post-Tariff and Community Supervision
Soar
wo inson
Ot
breaches of community
pre taritt post ten
ce aft Soper
(4 stopdown tepsown
aren, pet
area post tant cormunneraoaved
Stepeaide Srenaside endhg.
_—— , 0
stepacde eanacae Smee
tariff ending paroled community
supervision
Non DS°0 under
post tant a
fon esr community supervision
[I 3
ron DSPO non BSPD community
paroled supervision ending
non DSPD Retum
to Prison
For each of the routes through prison, it is important to differentiate between the part of
the sentence for which they are “pre-tariff’ (having time to serve before being eligible
for parole) and “post-tariff’ (having completed the minimum term, but still awaiting
release on parole which, for some, may never come). The time left “pre-tariff’ for those
coming off treatment programmes is calculated dynamically, because it will depend on
the length of waiting time that was experienced pre-programme, and that will have
varied according to the programme capacity and length of stay.
Men released into the community are subject to supervision either as part of their parole
conditions, or if these have expired and the offender is still high risk, under Multi-
agency Public Protection Arrangements. If they do not meet the terms of their licence
2 JG=
they may be recalled to prison. Community supervision is long-term (for life if that was
the sentence), and the model provides a count of the numbers released to community
supervision, differentiating between those who have been on a treatment programme
and those who have not. The model also simulates recidivism. People breaching their
licence return to the post-tariff stage in prison; those re-offending return to being newly-
sentenced.
There is undoubted scope to model community supervision in more detail.
Model Outputs
The main model outputs described the big picture of how many prisoners could be
expected at each stage (pre-programme, on programme, pre-tariff, post-tariff,
community-supervision) over the next 20 years, given assumptions about sentence rates,
and mean sentence length.
In addition, the client wanted to consider whether there would be enough special
treatment places, based on assumptions about the percentage of new prisoners of each
kind who would be suitable for treatment, programme capacity and length of stay.
The model structure itself posed interesting policy questions. For life-sentenced
prisoners, having a mean tariff length of 15 years, where the mean programme length is
5 years, most of the sentence will be spent not on the programme. If prisoners should go
on a treatment programme early in sentence, they will spend much longer in prison
post-programme. How should that time be spent? Can the gains made on the programme
be consolidated, or will they gradually wear off? Or would it be better to delay
admission to the programme until later in sentence, so that the treatment experience
might contribute towards preparation for release? Whilst it would be inappropriate to
comment in detail on the nature of these discussions, there is no doubt that such
questions for policy makers have a more potent force when posed by (or on behalf of)
the model.
As an outcome of the group model-building process, many such policy discussions were
triggered, not just about programme timing, but also about clarifying assumptions about
the intended purpose of the programme, its outcomes in terms of distribution of
prisoners across security categories, whether there should be separate assessment and
programme capacities (with two queues) or should assessment be a subset of
programme capacity, guaranteeing that following a successful assessment a prisoner can
proceed straight to the programme without delay.
Owing to the policy-sensitive nature of the project, actual numbers on some graphs are
not shown.
may We
Figure 4: Base Case Numbers of Prisoners of Each Kind
B® rar icontirr] 2: TotalPison{ite] 4: TotalPrison(Oet)
ee
it
Lapse
I
Lessing ——
ee ee ee ee
o zy cio 0 Fz)
age 2 Mints
2 Total in Pison by Setence Type
This graph shows the expected numbers of all life prisoners, all IPP prisoners and
“dangerous” determinately-sentenced prisoners on the twenty year period from 2002.
Figures are available for the previous forty years, but their only purpose is to set the
2002 distribution of prisoners across sentence-stages. There is a gradual increase in the
number of life-sentenced prisoners, consistent with policy assumptions about sentence
rates and average tariff lengths i.e. that tariff lengths are likely to get longer and
numbers being sentenced will slowly increase.
At the same time, it is assumed that courts will be using the new (IPP) sentence type to
dispose of those dangerous men found guilty of committing offences that previously
carried a determinate tariff. Over time, the number of dangerous men on determinate
sentences reduces, as the newly-sentenced are given indeterminate sentences.
Figure 5.1: Prison Programme Capacity (base case)
This graph shows the prison programme capacity, the total placed, and the total
numbers who are waiting for a programme place.
9B 1 pr dspd capa 2:0 prog 3:t0t wa pe
‘ha
;
;
| :
i
/ _
ee ee
fsa i
oe
1 if 2 Ww
i
Et of 3 fi
‘ee a ae oe 7b
Fae 1 onthe
2. tson rogamme Capacty, Occupancy, Numbers Wang
cae fe
The special treatment programme starts in year 2003, and the policy built-into the model
is to place only those in the early stages of sentence on the programme. The programme
starts with a capacity of 80 places, and grows at a rate of 10 places per year (shown as a
ramp-style increase rather than in annual steps). Based on this assumption, the
programme fills steadily reaching full capacity within three years. From then on, the
programme operates at full capacity, and with a steadily rising waiting list.
The model interface allows users to vary their assumptions about programme size and
maximum possible growth.
Figure 5.2: Prison Programme Capacity (increased)
In the example below, an annual rise of 20 over 14 years would be more likely to
provide sufficient capacity over the next 20 years, when another increase in capacity
would be indicated, other things being equal. As noted above, the number of prisoners
of this kind in the system is steadily increasing.
® 5: pr dspd capa 2: tot prprog 3: tot wait pr
‘|
ene
468 531 594 067 720
Page 1 Months
? Prison Programme Capacity, Occupancy, Numbers Waiting
Other possible solutions would include reducing programme length (but perhaps at the
expense of programme effectiveness in reducing risk) or increasing the threshold of
eligibility for treatment.
Similar considerations apply to the planning of hospital programme places.
2 JG=
Figure 6: Impact of Programme Effectiveness on Prison Population
As well as enabling the user to investigate these rather linear (not completely, because
of the impact of assessment and programme capacity) phenomena, it is possible to look
in more detail at the impact of the programme on the prison population. If the
programme is effective in reducing the level of dangerousness of its users, and they
consequently progress faster through to lower categories of security, then as more are
placed on the programme, the prison population is reduced On the other hand those in
the step aside group may stay for longer, hence increasing the prison population).
PP tating -2-3-4
=
a
is
on
ory zany obo ry cy
age 2 Months
2 Teal in Prison: IPP, Lite and “DSO Determinate”
The graph shows a comparison of the whole DSPD population in prison (IPP, Life and
Determinate combined) over four different model runs. In the first, there is no
programme capacity. In the second there is the “base case” capacity with annual
increases as described above. In the third, enough capacity is loaded into the system to
ensure that everyone who needs a treatment place gets it without any waiting. The final
run represents the same (effectively unlimited) capacity, with the added factor that
everyone completing the programme successfully completes it and enters the fastest
flow through to a lower security category.
For the current set of assumptions it is most striking that the programme has a very
limited effect on overall prison numbers. This is largely due to the long tariff lengths
that determine a large part of the overall length of stay in prison, combined with a risk-
averse representation of parole policy, where prisoners are held for a substantial time
even having reached the lowest security category. However, under some scenarios — e.g.
high assessment and low treatment availability — the prison numbers can be significantly
higher owing to more risky offenders being identified and staying longer in custody.
It also illustrates that the main purpose of the DSPD policy is not primarily treatment
with a view to reducing the size of the prison population but treatment as a part of a
system of public protection. The point is that those who are being released into the
community under supervision should have become less dangerous as a result of the
programme’’s impact.
2 OU =
Conclusions
This project provides a live example of a relatively brief project to support public policy
making using group model-building. It contains findings that are relevant to group
model building, as well as providing an example of a model representing the main
dynamics of treatment programmes within prison.
The success of the group model building component is easy to recognise but harder to
measure. Although various formal and informal groupings were involved at different
stages of the modelling project, there was always a representative of the core policy
staff group present at each of these meetings. By this means the core group retained a
sense of ownership of the model structure, and were content with the main stock-flow
representations and feedback loops.
In the second iteration, the staff group were more assertive in describing the stock-flow
structure, requiring facilitation, rather than direction, from the modeller, along the lines
of: “we see things this way; how can we represent this best?” The manager, having
knowledge of management science but not specifically system dynamics, began to work
directly with the Ithink software, and other members of the team gained hands-on
experience, more at the level of formatting diagrams rather than model-building.
Towards the end of the project, most of the core group underwent four days of training
in system dynamics using this software, based around this model.
In the experience of the modelling team, that represents a more detailed engagement
with system dynamics concepts and constructs than is normally achieved in projects of
this brevity.
The main outcomes are:-
o auseful map in the form of a stock-flow diagram of the prison / hospital / special
programme treatment system, which itself triggered fruitful policy discussions
o some broad estimates of the likely numbers of prisoners who will be flowing
through this system over the next 20 years
© an approach to capacity planning in which capacity can be added in a number of
ways, and its impact shown on the whole system, largely through the effects of
having large numbers waiting, some of whom will not benefit from the special
treatment programme
o. the beginnings of an understanding of how system dynamics might bring useful
insights in planning other aspects of the criminal justice system
mee fee
References
Hirsch G, Homer J, McDonnell G and Milstein B, 2005. Achieving Health Care Reform
in the United States: Towards a Whole System Understanding, Paper presented at the
23™ International Conference of the System Dynamics Society, Boston, USA.
Luna-Reyes LF, Martinez-Moyano IJ, Pardo TA, Cresswell A, Andersen D, Richardson
GP. 2006. Anatomy of a group model-building intervention: building dynamic theory
from case study research. System Dynamics Review 22(4): 291-320
Rouette AJA, Jongebreur W, van Hoof P, Heijmen T, Vennix JAM. 2004. Modelling
Crime Control in the Netherlands, Proceedings of the 22"! International Conference of
the System Dynamics Society
Wolstenholme EF, Monk D, Smith G and McKelvie D. 2004a. Using System Dynamics
to Influence and Interpret Health and Social Care Policy in the UK.
Proceedings of the 22" International Conference of the System Dynamics Society
Oxford, England.
Wolstenholme EF, Monk D, Smith G and McKelvie D. 2004b. Using System Dynamics
in Modelling Health and Social Care Commissioning in the UK, Proceedings of the 22
International Conference of the System Dynamics Society Oxford, England.
Wolstenholme EF, Monk D, Smith G and McKelvie D. 2004c. Using System Dynamics
in Modelling Mental Health Issues in the UK, Proceedings of the 22" International
Conference of the System Dynamics Society Oxford, England.
Wolstenholme EF, Repper D, Monk D, Todd D and McKelvie D. 2006. Reforming
Mental Health Services in the UK - Using System Dynamics to support the Design and
Implementation of a Stepped Care approach to Depression in North West England,
Proceedings of the 24" International Conference of the System Dynamics Society
Nijmegen, Netherlands
Vennix, JAM and Gubbels JW. 1992. Knowledge Elicitation in Conceptual Model
Building : A Case Study in Modeling a Regional Dutch Health Care System, European
Journal of Operations Research 59(1): 85-101.
Vennix JAM. 1996. Group Model Building, Facilitating Team Learning using system
dynamics, Wiley London.
oD xc
