Utilising buffer management to manage patient flow

Abstract

TOC buffer management has been successfully developed to improve patient flow across emergency and planned health and social care within the UK and Holland. This paper details these developments before attempting to uncover the theoretical basis of the approach through multi-case research across three UK hospitals. The authors demonstrate how buffer management concepts can be used to emulate the kanban control principles in a health and social care environment and thereby strategically support lean implementations.

Full text

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
Utilising buffer management to manage patient flow
Roy Stratton
*
and Alex Knight
Nottingham Business School, Nottingham Trent University, Burton Street, Nottingham,
NG1 4BU, UK (E-mail: roy.stratton@ntu.ac.uk)
*
QFI Consulting, POBox 953, Tring, Herts, HP23 4ZX, UK
Abstract
TOC buffer management has been successfully developed to improve patient flow
across emergency and planned health and social care within the UK and Holland. This
paper details these developments before attempting to uncover the theoretical basis of
the approach through multi-case research across three UK hospitals. The authors
demonstrate how buffer management concepts can be used to emulate the kanban
control principles in a health and social care environment and thereby strategically
support lean implementations.
Keywords: Theory of Constraints, kanban, health care
The operations challenge in healthcare and social care
With an aging population it is increasingly important to view health care delivery
systems holistically, acknowledging the wider delivery system encompasses both health
and social care. Therefore, effective and efficient acute heath care must also ensure
strategic alignment of the social care system. This fact further demonstrates the
complexity of this wider system as many different organisations need to be
synchronised to ensure improvements in efficiency and timeliness associated with
improved patient flow. However, the silo mentality, driven by separate organisational
targets, often results in dysfunctional behaviour. In the UK the government drive to
improve performance in
the National Health
Service (NHS) has had to
address many examples of
such behaviour, especially
at the silo interface
between social and
medical care services.
Figure 1 illustrates the
macro flow routes and the
interface between health
and social care. This
figure, in addition to the
major flows, shows the
variation and uncertainty
ED
Acute CH
So cial &
H ealth
C are
Home
Social &
Healt h
C are
Social &
Health
Care
Social &
Health
Care
Hours
Ho urs
GP
referrals
Ambulances
Minors
Home
Home
Home
Days
Days
Days
Days
Days
Electives
Resi dential
& Nursing
Care
Figure 1 HEALTH AND SOCIAL CARE SYSTEM
Figure 1 HEALTH AND SOCIAL CARE SYSTEM
-
-
The chain of activities
The chain of activities
ED: Emergency Department CH: Community HospitalGP: General Practitioner

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
associated with demand, process and recovery times. In the UK such times together
with budgetary control are used to drive improvement and there are now emergency and
planned treatment targets. The Emergency Department (ED) performance is measured
against a government target for England that states 98% of patients are to be treated
within 4 hours. The more recent planned treatment target states that 98% of patients are
to have started treatment within 18weeks of referral by a General Practitioner (GP).
In the UK the government sets the waiting targets and the local management are
empowered to find appropriate means of achieving these targets whist remaining within
budget. Agencies, such as the NHS Institute for Innovation and Improvement support
this process and over the years a wide mix of tools and techniques have been adopted
but in the last three years this has increasingly centred on lean thinking.
This would appear to mirror worldwide healthcare interest in lean that is reflected in
the growth in lean healthcare books (Zidel, 2006; Fillingham, 2008 and Graban, 2009).
As a consequence a flow focus has emerged, with an emphasis on eliminating wasteful
variation associated with batching, poor layout discipline and mis-synchronisation.
There is a clear emphasis on continuous improvement with much of the lean activity
centring on kaizen rapid improvement events that typically last up to a week. However,
these kaizen events are often not strategically focused as suggested by Dan Jones in his
Lean Enterprise August 08 news letter.
‘Improvement is not so easy to sustain… With only loose direction from top
management it is difficult to trace the results from these islands of improvement…
and no one checks. We recently did an assessment of a hospital that had done 93
kaizen events. The success rate was less than 20% and none of these had impacted
the core A&E process that really kept the CE awake at night’ (Jones, 2008)
The traditional means of strategically focusing continuous improvement in Ohno’s
Toyota Production System is through kanban control (Ohno, 1988: 30).
‘Kanban is a way to achieve just-in-time; its purpose is just-in-time. Based on this,
production workers start work by themselves, and make their own decisions
concerning overtime. The kanban system also makes clear what must be done by
managers and supervisors. This unquestionably promotes improvement in both work
and equipment.’ (Ohno, 1988:29)
However, kanban or its equivalent does not feature in the application of lean
thinking to healthcare. This paper argues that a variant on TOC buffer management is
effectively meeting this requirement in the management of patient flow in both
emergency and planned care. QFI Consulting have led these developments within the
UK NHS over more than five years which have been highly regarded by the senior
management of the hospitals concerned as indicated by the following quotes.
‘TOC has been applied to improved patient flow in A&E, Assessment Units, and
discharge planning. This has resulted in a sustained reduction in medical
length of stay from 8.6 to 6.3 days (>25%). Released bed capacity supported the
achievement of the 18 week GP referral to treatment target, a year ahead of
schedule.’ Martine Price, Director of Governance and Nursing, Taunton and
Somerset NHS Foundation Trust (October 08)

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
‘With the help of Theory of Constraint we have been able to move Barnet &
Chase Farm Hospitals NHS Trust from one of the worst performing trusts in
England to one of the top performing. In Q4 (2007-2008) we were the top
performing trust in London for the 4 hour target and 6th across England. Also,
by applying the Theory of Constraints to our discharge process we have been
able to reduce our length of stay by 27% and we know we can improve further
on this.’ Averil Dongworth – CEO Barnet and Chase Farm NHS Trust (October
08)
‘The application of TOC has helped us to reduce our length of stay by up to 23%
in one of our hospitals, but the real benefits from QFI Jonah are around
improving how we deliver care to our patients through better planning and co-
ordination of their care - ultimately it's about recognising that patients should
go home as quickly and safely as possible.’ Jeffrey Worrall, Chief Executive,
Sherwood Forest Hospitals NHS Foundation Trust (October 2008)
This research has, therefore, been conceived to help explain the nature and extent of
the improved performance in relation to the following research questions.
Research question
• How and why has buffer management contributed to improved patient flow?
• How does buffer management relate to kanban control?
Research method
This case research involved access to the buffer management applications, the
implementation process and the users of the system. Archival data, reports and semi-
structured interviews were used to answer the research questions in line with the good
practice advocated by Yin (1994) and Eisenhardt (1989). A mix of hospital and social
care staff were involved in gathering relevant data over a period of 18 months.
Buffer management – origins and traditional applications
Whereas Ohno (1988) developed kanban control to support the management of
inventory flow (Ford, 1928), Goldratt focused on make to order (MTO) (1990) and
project environments (1997) in developing the concept of time buffers. In the case of
MTO the overall approach is termed Drum – Buffer - Rope (DBR). This approach is
well documented and has been developed as a practical approach in the manufacturing
environment (Goldratt, 1990; Umble and Srikanth, 1997; Schragenheim and Detmer,
2001; Stratton et al., 2008) as detailed below.
• Drum: The drum can be a resource or market demand and we will assume for
comparison it is market demand here.
• Rope: This is the planning mechanism for releasing work and effectively chokes
material release in line with customer requirements. This is the DBR mechanism
for preventing over production where the rope represents the time offset that
releases material in line with demand.
• Buffer: Once material is released the time remaining is termed the buffer time,
based on the assumption that touch time is negligible. To support the management
of this buffer it is simply divided into three equal zones (See Figure 2).

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
The four main functions of DBR buffer management
1) Prioritise the flow of work according to buffer penetration.
Work is release based on the choked release date set by the rope. Work is progressed
with reference to the buffer penetration indicated either by the colour coding or a buffer
penetration percentage.
2) Identify when to expedite potentially late arrivals.
The red zone should effectively represent the tail of the distribution and relatively few
orders should enter this zone
(see Figure 2). If they do
this signals the need to act,
that is, expedite where
necessary to ensure it is
delivered within the
remaining buffer. As has
already been stated even
though the red zone
represents only one third of
the lead time there is still
time to complete an order
even if it has not even been
started by this time.
3) Signals when the whole
production system is
starting to become unstable.
The system is acknowledged to have the capability to expedite a limited number of parts
but if that rises above a certain level (Goldratt (1990) suggests over 10%) the system is
at risk of going out of control. In such cases urgent action is required to bring the system
back into control typically by limiting order intake or increasing capacity temporarily.
4) Identify main sources of delay in order to target improvement.
Continuous improvement is about reducing variation (Deming, 1986) and in the flow
context targeting sources of variation that threaten delivery, as with the kanban ‘river
and rocks’ analogy. The equivalent in buffer management is associated with the
exposure of the red zone and, therefore, it is necessary to record in some way why the
order has entered the red zone. This information is periodically analysed using Pareto
analysis and used to target the cause using lean tools.
Project based buffer management
In project management the assumption that touch time is a small percentage of the lead
time is rarely valid. In this environment queuing, as with operations, does not dominate
and, therefore, touch time, as in the task times, is a significant part of the overall lead
time. Goldratt (1997) suggests the touch time is typically around 50% of the lead time.
However, in this environment variation and uncertainty predominate and it is imperative
that the associated time buffering is effectively managed to improve flow. In the
traditional planning of projects, activity duration estimates are commonly formed into
what is referred to as the critical path method. As with production the lead time
including buffer time is allocated to each task in an attempt to accommodate the task
uncertainty. However, informally allocating this ‘time buffer’ at the task level neither
protects the project delivery promises effectively nor minimise the lead time.
The TOC approach to buffer management in projects is called Critical Chain Project
Lead time
Probability of completion
Green Amber Red Black
Figure 2:Buffer management in an operations environment showing
buffer zoning of lead time distribution
Rope

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
Management (CCPM) (Goldratt, 1997; Steyn, 2000; Stratton 2009) and as with DBR
emphasises the need to manage the time buffer to accommodate and reduce the
variation and uncertainty. However, in this case the touch time is a significant part of
the lead time and the buffer concept needs to be redesigned to suit the project
management application. To address this, the CCPM buffer is placed at the end of the
project in what is called a project buffer and for non-critical activities there is a lower
priority feeding buffer between that non-critical dependent chains and the critical chain.
To create such buffer the excess time needs to be extracted from the task estimates.
Goldratt (2007) advocates the project buffer and associated feeder buffers should be set
at one third of the lead time of the related activities. With the buffers now positioned to
protect the project as a whole the overall lead time arguably can be reduced. In
summary there are three prime functions of CCPM buffer management.
The four main functions of CCPM buffer management
1) Provide a priority mechanism across projects base on buffer status.
A simple ratio of project buffer remaining to critical chain remaining enables resource
providers to prioritise across projects.
2) Provide a mechanism for protecting the buffer and escalating recovery action.
Unlike with DBR where action is only taken in the red zone it is important to reduce
buffer consumption at all stages of a project. As there is effectively only one activity
consuming the project buffer at any time the improved visibility of buffer consumption
creates awareness and opportunity to support the resource concerned.
3) Track the prime causes of buffer consumption to target ongoing improvement
activity. This is more complex than in DBR as the cumulative causes of delay are
required and the relative importance is only subsequently known. Such systems are only
feasible through computer support.
4) As with production there is a need to control the release of new projects to ensure
lead times do not unnecessarily grow through multi tasking (Goldratt, 2007). This is
typically achieved by choosing a resource towards the end of the project and ensuring
the number of live projects is controlled.
Emergency and planned care buffer management applications
In healthcare two hybrid buffer management systems have emerged through the work of
QFI Consulting as detailed below.
Emergency Department buffer management application
The ED embodies certain characteristics that need to be mapped onto the DBR approach
in establishing the validity of the approach and how it can be effectively applied to this
environment. The means by which QFI have achieved this is outline below. The design
centres on the NHS 4 hour treatment target.
• Drum: The pace of the system is set by the drum and in this environment it is
determined by demand with no prior warning apart from the ability to forecast. The
Department is expected to manage capacity in line with demand.
• Rope: The rope would normally be used to control release of work into the system
but in this environment there is no attempt to control release as all patients are
considered to be in the 4 hour buffer on arrival.
• Buffer: The buffer time of 4 hours commences on patient entry and as with
standard buffer management practice this time is equally divided into three zones (see
Figure 2) of 80 minutes each. This buffer management application is supported by

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
software that tracks the patient buffer consumption with the priority being set by the
time of entry. The buffer changes colour as time progresses and at each transition the
activity currently delaying progress is selected from a drop down menu to provide data
for subsequent analysis. This screen (see Figure 3) is updated by the person responsible
for the patient and on discharge, or transfer to a Medical Assessment Unit (MAU), the
patient is automatically removed from the screen. The practical application of the four
functions of DBR buffer management are as follows:
1) The order of priority regarding the discharge target is simply the order of
presentation, however, this clearly does not acknowledge clinical priorities that are
expected to override this default priority sequence.
2) Patients in the red are within 80 minutes of breaching the target and the networked
screen is used across the hospital to help expedite action by the resource causing the
delay. The resource causing the delay of a patient is also informed separately from this
screen.
Figure 3 ED Buffer Management system screen
3) The overall stability of the Emergency Department performance is monitored by the
number of red zone penetrations and middle and senior management have visibility of
this screen and can assess the situation and allocate additional resource to avoid ED
going unstable. On an hourly basis analysis of the resources causing most delay across
all patients is communicated. This escalation is deemed to be an essential part of the
synchronisation of resources to enable flow.
4) The reasons for delay are subsequently analysed by charting the main causes of
delay on entering the yellow, red and back zones. This data is used in buffer
management meetings (typically weekly) aimed at focused continuous improvement
effort. A Pareto graph is produced to present the main reasons for delay but this is also
used in looking for causal links and any common causes that become a focus for rapid
improvement.
Patient discharge buffer management application
Timely patient discharge applies to acute and community hospitals as well as social care
homes and can be viewed as key to patient flow through the health and social care
system. This application aims to shorten the length of stay (LOS) by subordinating all

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
support activities to the medical needs of the patient with the aim that the patient’s
position within the health and social care path is determined by medical needs alone.
A first step in this discharge planning process is to set a Planned Discharge Date
(PDD) when the patient first arrives as well as identify the activities that need to be
completed before this date. In reality the PDD is initially indicative but can be updated
as knowledge and circumstances dictate. The concept of a PDD is new to many
hospitals but is essential to the planning and control system and as with CCPM the
emphasis is on regularly updating the PDD to ensure relative priorities reflect the
medical needs.
With the time to discharge based on medical judgement alone it is necessary to ensure
subordination (alignment) of the support activities. In this environment activity
durations are more characteristic of a project where the touch time is significant as
illustrated in Figure 4. Therefore, the projected buffer time remaining comprises the
time to the planned discharge date (PDD) minus the longest activity duration yet to be
completed.
Accounting for the duration of activities has been simplified with the assumption that
there is no activity dependency, therefore, the longest activity duration together with the
time now and PDD enables the extent of the projected buffer penetration (see Figure 4).
This application is effectively a hybrid combination of the CCPM and DBR buffering
approaches. That is, the
duration to PDD
effectively acts as a
buffer as in DBR but the
longest activity duration
is acknowledged in
determining the buffer
penetration. Finally, the
buffer regions have been
adjusted (see Figure 4) to
account for the activity
duration offset. The
software can provide
priority lists for the
different specialists as in
the case for continuing
care.
Buffer management
Four buffer management functions operate in Discharge Buffer Management:
1) The priority status unlike ED will change with changes to the PDD and any changes
to the longest remaining activity to complete. As PDDs can change in both directions
the relative priority of different patient activities will change and are displayed simply
through the on screen colour code.
2) Patient projected to stay beyond the PDD are shown in the black, therefore
urgent, but as this predicts a future situation there is some room for recovery. The
‘reasons for delay’ data is circulated daily and the main reason for delay communicated
as part of the escalation process.
3) As with DBR buffer management, the stability of the overall system is signified by
the overall number of red and black buffer penetrations which provide advanced
warning of the pending instability and the need to escalate recovery action.
Patient
Arrives
Planned
Date of
Discharge
0
5 10
50% 25%
25%
Continuing Care Form
(3 days)
Occupational
Therapy (2 days)
Physiotherapy
(2 days)
• The current time is in
the green zone but the
patient is projected to
be in the red due to an
activity duration of 3
days
Days
Figure 4 Discharge Buffer Management
Activities

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
4) As with DBR
buffer
management the
tracking of the
activities that
cause the delay
are recorded to
support the
escalation process
where the impact
of the ‘reason for
delay’ is measured
but also analysed
to identify
common causes
(see Figure 5).
Figure 5 Discharge QFI Jonah top delay reasons by region
These are then analysed at buffer management meetings, at the ward and the
hospital/region level.
Discussion
The research evidence supports the claims of the CEOs that buffer management
implementations having rapid, significant and a sustainable impact on patient flow, as
illustrated in Figure 6.
Figure 6: Improvement in emergency breaches following buffer management
implementations introduced from March to July 08.
These improvements are the result of introducing both ED and Discharge Jonah buffer
management within the wider delivery system. It is evident that understanding the flow
relationship from ED through to social care has proved to be centrally important to the
success of this approach which embraces these key concepts within a simple but
No of Breaches against 4 hour Target: March - August
0
200
400
600
800
1000
1200
02/03/2008
16/03/2008
30/03/2008
13/04/2008
27/04/2008
11/05/2008
25/05/2008
08/06/2008
22/06/2008
06/07/2008
20/07/2008
Weeks
No. of breaches
BHRT
KGH
QH

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
effective variant on more traditional buffer management.
Buffer management is concerned with managing flow in a variable and uncertain
environment which clearly has links to earlier seminal work. Shewart (1931; 1939)
introduced the modern concept of continuously reducing sources of variation and the
associated waste, but it was Ford (1928) and then Ohno (1988) that showed how this
applied in a flow environment. Buffer management would appear to conceptually link
Statistical Process Control (SPC) and flow management, thereby providing means of
managing the impact of variation and uncertainty on patient flow. Whereas Ohno (1988)
used physical inventory buffering and the concept of enforced problem solving to target
improvement time would appear to be a more appropriate choice in the healthcare
environment.
The buffer zoning, as with SPC, provides a simple means of interpreting the
instability signals, focusing timely expediting and continuous improvement effort.
Buffer management also has the potential to provide a timely signal to ensure growing
capacity issues are escalated as the number of red zone penetrations are seen to change.
This is the focus of some on-going research into establishing a responsive capability to
dynamically adjusting capacity in response to special causes, such as trends in demand.
The authors experience is that the buffer management concept is deceptively simple
and it is easy to overlook the underlying concepts that make it work. This is evident in
the growing number of commercial software tools that are not methodologically driven
in their design and use.
Conclusions
Goldratt’s buffer management has provided an effective means of extending the flow
principles associated with Ford and Ohno’s work to patient flow in health and social
care. The focus on time, as oppose to inventory, has enabled the complex nature of
patient flow to be controlled though the use of buffer aggregation and zoning, akin to
Shewart’s SPC application. As with SPC this enables both system control signalling and
the targeting of improvement activity. This paper has demonstrated how the concepts
and theory has been practically extended to this environment as well as demonstrating
the potential of integrating lean and TOC developments to improve overall patient flow.
Whareas lean is acknowledged to be a medium to long term approach (Jones and
Mitchell, 2006:3) TOC buffer management has demonstrated significant short term
improvement to patient flow. The authors intend to follow this paper with a more
detailed study of how lean and TOC buffer management can be effectively combined as
advocated by Jones (2008).
‘The next phase of lean starts with top management learning to see the vital
few actions, on the right products and service lines and the right customers
that would have the biggest impact on the performance of the organisation.
In the hospital getting the chaotic flow of patients through A&E under
control is the key to establishing stability and a common rhythm across the
rest of the hospital.’ (Jones, 2008)
The authors respectfully propose that buffer management is a natural successor to
kanban that enables flow control to be realised in a health and social care environment.
However, more in depth research is needed to explore the underlying principles and
develop their practical application further.
References
Bhasin, S. and Burcher, P.(2006), “Lean viewed as a philosophy,” Journal of

Proceedings of the 16
th
International Annual EurOMA Conference (Implementation –
realizing Operations Management), Goteborg, Sweden, June 09
Manufacturing Technology Management, Vol.17, pp. 56-72.
Deming, W.E. (1986), Out of the Crisis. Cambridge, MA: MIT.
Eisenhardt, K. M. (1989), Building Theories from Case Study Research. Academy of
Management Review, 14(4), pp532-550.
Fillingham, D., (2008), Lean healthcare. Kingsham Press, UK.
Ford, H. (1926), Henry Ford – Today and Tomorrow. Portland, OR:
Productivity.
Goldratt, E.M. and Cox, J. (1984), The Goal. North River Press, NY.
Goldratt, E.M. (1990), Theory of Constraints. New York: North River Press
Goldratt, E.M. (1997), Critical Chain. North River Press, Great Barrington, MA.
Goldratt, E.M. (2007), The Goldratt Webcast Program on Project Management Viewer
Notes. V 4.71. GMG: Amsterdam.
Graban, M., 2009, Lean Hospitals, Productivity Press, NY.
Jones, T.J. (2008), Lean Enterprise Academy News Letter – 18 August. Download from
www.leanuk.org on 13
th
Nov 08.
Jones, D., Mitchell, A. (2006), Lean Thinking for the NHS, NHS Confederation,
London.
Ohno, T. (1988), The Toyota Production System; Beyond Large-Scale Production.
Portland, OR: Productivity Press.
Schragenheim, E., and Detmer, W. (2001), Manufacturing at Warp Speed. CRC Press.
Shewart, W.A. (1931), Economic Control of Quality of Manufactured Product. New
York, NY: Van Nostrand.
Shewart, W.A. (1939), Statistical Method from the viewpoint of Quality Control.
Washington. DC: Graduate School of the Department of Agriculture.
Steyn, H. (2000), “An investigation into the fundamentals of critical chain project
scheduling,” International Journal of Project Management 19, pp363-369.
Stratton, R., Knight, A., and Williams, M. (2005), “Theory of Constraints and
Healthcare Management”, Proceedings of the international conference of
EUROMA (Operations and Global Competitiveness), Budapest, pp 1989-1998.
Stratton, R., Robey D. and Allison I. (2008), “Utilising Buffer Management to Manage
Uncertainty and Focus Improvement,” Proceedings of the International Annual
Conference of EurOMA, Gronegen, the Netherlands (June).
Stratton, R. (2009), “Critical Chain Project Management – Theory and Practice,”
Proceedings of the 20
th
POMS Conference, Orlando, May 1
st
-4
th.
Umble, M.M., and Srikanth, M.L.,(1997), Synchronous Management: Profit
based manufacturing for the 21
st
Century, Volume two: Implementation Issues
and Case Studies. Spectrum Publishing Company, Guilford, CT.
Umble, M.M. and Umble, E.J. (2006), “Utilising buffer management to improve
performance in a healthcare environment,” European Journal of Operational
Research 174, pp 1060-1075.
Womack, J.P. and Jones, D.T. (1996), Lean Thinking-banish waste and create wealth
in your Corporation, New York: Simon and Schuster.
Yin, R.K. (1994), Case Study Research – Design and Methods. 2
nd
Ed., London: SAGE.
Zidel, T.G. (2006), A Lean Guide To Transforming Healthcare, ASQ Quality Press,
Milwaukee, Wisconsin.