leading edgeOCTOBER 2005 ISSUE 10

The potential of systemdynamicsA new era of strategic planning?

Key points

To make significant changes incomplex adaptive organisations itis necessary to think differentlyand test ideas before use.

System dynamics:

• is the origin of the current trend of ‘whole systems thinking’

• enables the strategic modelling of non-linear relationshipswithin complex adaptiveorganisations such as the NHS

• works along pathways of carewhich cross organisation andsector boundaries

• complements existing planningtools such as spreadsheetmodels or actuarial analysis.

System dynamics is a versatile way of bringing whole-systemsthinking to life in a rigorous,testable way.

System dynamics is a way of thinking about the future which focuses on‘stocks’ and ‘flows’ within processes and the relationships between them. Itcan facilitate ideas for both specific solutions and generic ‘new world’ rules.It is a risk-free way of refining plans before implementation, and of testingideas using computer simulation that follow patient pathways and crosshealth and social care organisational boundaries.

System dynamics goes beyond the scope of spreadsheets, process maps anddata analysis to challenge existing thinking and help design, test andcommunicate sustainable solutions. It is currently being used in more than15 local health and social care communities in the UK to simulate novel waysof achieving win–win solutions, make assumptions more explicit, andidentify inconsistencies between data, process and policy.

This Leading edge briefing outlines the background to, the thinking behindand the toolset of system dynamics. It describes how it is applied, examinescase studies and outlines its limitations. This paper was prepared byProfessor E Wolstenholme of Symmetric SD, recipient of the 2004 Jay WrightForrester Award for the best contribution to system dynamics in the last fiveyears. It also draws on the work of the Whole System Partnership.

importantly, ‘dynamic’ complexity(having many interconnections). They have long processes whichoften cross organisational boundariesand have their own cultures,accounting systems and performancemeasures. Decision-making in the

The challenge facinghealthcare organisations

Healthcare organisations are large,complex and adaptive. They showboth ‘detailed’ complexity (containinga large number of elements) and,

2

been static in nature, sector-basedand reliant on data and spreadsheets,with limited transparency ofassumptions. To make significantchanges in such organisations it isnecessary tothink differently and testideas before use as effective solutionsmay often seem to be counterintuitive and even defy logic. Systemdynamics provides a method oftesting solutions.

What is system dynamics?

System dynamics was conceived atMIT in the late 1960s. It has nowgrown into a major discipline(Sterman, 2000) that is widely used inthe private sector in, for example, oil,asset management, financial services,defence and consultancy. Its basis islearning to see patterns of behaviourin organisations and grounding thesein the structure of organisations –their operational processes andpolicies. It uses software to mapprocesses and policies at a strategiclevel, populate the map with data,and simulate the evolution of theprocesses under transparentassumptions, policies and scenarios.

leading edge 10 The potential of system dynamics

public sector is made more difficultby organisations having differentplanning time horizons.

Because of our limited ability tounderstand the multipleinterconnections, we find it difficultto manage complex adaptiveorganisations. Managers tend todevelop mental models linkingstructures, processes, patterns andevents, and build up a picture of howorganisations work which may notreflect reality (see Figure 1.) Incomplex adaptive organisations this can result in unintendedconsequences that undermine well-intended actions, and may leadto many types of informal copingactions which in turn inhibitimprovements.

Traditionally, solutions to problems incomplex adaptive organisations have

Figure 1. Systemic hierarchy

‘In the near future I predict that allinvestment planning projects willbe subjected to dynamic appraisalfor their effects on the wholepatient pathway.’

Steve Arnold, 2004

System dynamics is the origin of thecurrent trend of ‘whole systems thinking’in health and social care. It provides aset of thinking skills and a set ofmodelling tools, as described below.

Thinking skills

A wide range of thinking skills andabilities are required to understandcomplex adaptive organisations.These include (Richmond, 1998):

• dynamic thinking –conceptualising how organisationsbehave over time and how wewould like them to behave

• ‘system-as-cause’ thinking –determining plausible explanationsfor the behaviour of theorganisation over time in terms ofpast actions

• ‘forest’ thinking – seeing the ‘bigpicture’ and transcendingorganisational boundaries

• operational thinking – analysingthe contribution which differentoperational factors make to overallbehaviour

• ‘closed-loop’ thinking – analysing‘feedback loops’, including the waythat results can influence causes

• quantitative thinking –determining the mathematicalrelationships needed to modelcause and effect

• scientific thinking – using modelsto construct and test hypotheses.

Modelling tools

System dynamics is a strategic ratherthan an operational tool. It can beused to integrate policies acrossorganisations where behaviouralfeedback is important, and analyse

EventsToday’s pressure, for example

waiting lists

View of world – what’s important

Patterns

Mental models

Structures –processes/policies

A

A

CB

today

3

variation. The key elements of systemdynamics are shown below.

Stocks and flowsProcesses in system dynamics areviewed in terms of ‘stocks’ and ‘flows’.Stocks are measurable accumulationsof physical (and non-physical)resources, for example patients,customers and work backlogs.Essentially they are delays whichseparate and buffer their inputs from their outputs, and are built up and depleted over time asinput and output rates into themchange. They persist through timeand continue to influence other flowseven when their inputs and outputsare switched off. Recognising thedifference between stocks and flows is fundamental tounderstanding systems.

Organisational boundariesRecognising organisational sectorswith different cultures is key to

understanding the impact ofboundaries on patient pathways.

Learning through modellingSystem dynamics improves decision-making by extending andsharing mental models acrossmanagement teams. It demands skillin conceptualisation. Figure 2 shows the process of applyingsystem dynamics.

Comparison ofspreadsheets and systemdynamics models

Operational tools (such as processmapping, spreadsheets, data analysisand discrete event simulation) havecomplementary roles in the detailedoperational planning for sections ofthe patient pathway. For example,spreadsheets support quantitativethinking but can only be used forsimple simulations. In contrast,

system dynamics facilitates a moresophisticated, quantitative simulationcapable of more robust and reliableoutcomes. A comparison ofspreadsheets and system dynamicsmodels is shown in Figure 3.

How is system dynamicsused?

Step 1. System dynamics requires asignificant managerial issue and acommitted, consistent managementteam representing all agenciesinvolved in the issue. ‘Translators’,experienced in both conceptualisingand formulating system dynamicsmodels, are required to work with themanagement team.

leading edge 10 The potential of system dynamics

‘Recognising the difference betweenstocks and flows is fundamental tounderstanding systems’

Figure 2. The application of system dynamics

Source: Symmetric SD training material 2005

Managementteam

Policyexperiment

Structure v

behaviour

IssueBehaviour over

time graphs

Stock-flow maps andsimulation models

Data

now

PI ‘A’

PI

Time

CObjective

Mentalmodels

Visions

StrategiesCausal maps

AB

D

E

4

Step 2. Trends in the majorperformance measures of theorganisation, and their futuretrajectories – both desired andundesired – are analysed.

Step 3. The key contribution ofsystem dynamics is then to formulatea high-level process map linkingprocesses across organisations andpopulated with the best data available.

Step 4. They must be capable, whensimulated, of demonstrating theirability to match the trends of Step 2 (‘what is’ analysis). At thisstage, the rigour involved inknowledge-capture and indeveloping the quantitativesimulation model can identifyimportant mismatches andinconsistencies between howmanagers claim their organisationswork and observed data, patientpathway descriptions andperformance outcomes.

Step 5. Once validated, simulationmodels can then be used effectivelyto test radically new ideas (‘what if ’analysis).

As an example of system dynamicsat the highest level, the whole ofhealth treatment and illnessprevention can be shown as a stockand flow map. This health well-beingprocess is shown in Figure 4 and

Figure 5 amplifies this process toinclude factors which affect thestocks and flows of people throughthe system. The main messagecommunicated by such a map isthat delays in upstream effort (illnessprevention) reinforce downstreamaffliction and dependency onprofessionals, which furtherundermines upstream effort.

Case study: delayed hospital discharges

System dynamics has been used toinfluence health and social carepolicy (Wolstenholme, 2004). During2002 the Local GovernmentAssociation (LGA) and the NHSConfederation commissioned asystem dynamics model fromSymmetric SD (some of whom wereworking for OLM and Cognitus atthat time) to address theGovernment’s plans for a CommunityCare (Delayed Hospital Discharge) Bill.The model was intended to underpina national campaign to influenceGovernment policy and prevent theintroduction of ‘fines’ on socialservices.

The model was developed with agroup of NHS and social servicesdepartment managers, using national

leading edge 10 The potential of system dynamics

Figure 3. Comparison of spreadsheets and systemdynamics models

Spreadsheet models

Many external inputs over time

Does not recogniseaccumulations/stocks

Input interface is matrix of cells

All calculations are instantaneous

Opaque – ‘black box’

Applied by individual analysts

Relationships between cells largelylinear

Does not use sectors ororganisational boundaries

Easy to create and use

System dynamics models

Each time point calculated internallyon information feedback and policy

Processes explicitly represented asstock accumulations and flowsbetween ‘sources’ and ‘sinks’

Input interface is map of physicalstructures

Calculations recognise delaysinherent in stock structure, sooutput is phased, as in reality

Transparent

Shared by management teams

Relationships often non-linear

Represents sectors and organisationboundaries

Challenging to create and use well

Figure 4. The health well-being process

Source: (Hirsch, 2005)

Dying fromComplications

Becoming vulnerable

DevelopingComplications

Becoming Afflicted

Becoming no longervulnerable

VulnerablePopulationGeneral

Population

Afflicted without

Complications

Afflicted with

Complications

5

data to simulate pressures in asample health economy coveringprimary, acute and post-acute careover a three-year period. The modelwas driven by variable demand,including three winter pressure‘peaks’ when capacity in each sectorwas stretched to the limit. Itincorporated a number ofmechanisms by which agenciescoped during periods of highdemand – for example, movingmedical patients to surgical beds andearly discharges with allowance forre-admissions. Figure 6 gives anoverview of the sectors of the model.

The obvious solution of introducingmore hospital capacity in the acutesector to relieve problems wasshown to be a classical ‘fix that fails’(Wolstenholme, 2003), increasingacute admissions but alsocompounding delayed discharges.

leading edge 10 The potential of system dynamics

The model illustrated that, despitethe small numbers of patientsrequiring post-acute care, thecounter-intuitive, alternative policy ofincreasing post-acute capacity couldprovide a win–win solution for bothacute and post-acute performancemeasures. Not only werereimbursement charges low, but byproviding relief from the effects ofcoping actions the acute sectorwould reap the rewards of both lowelective waiting times and higherelective activity.

This model generated considerablepolitical interest. It was, apparently,instrumental in causing somerethinking of the intended legislationso that social services were provided with investment funding toaddress capacity issues and the ‘re-imbursement’ implementationwas delayed for a year.

Case study: Using a healthsystem simulator to testworkforce planning

During 2004, Shropshire andStaffordshire Strategic HealthAuthority worked with the WholeSystems Partnership to improve the evidence base and intelligenceunderlying strategic investmentdecisions, in terms of both physical capacity – beds, day case capacity, outpatients,intermediate care etc – and thelikely impact on the workforce,particularly in terms of skill-mix and location.

The modelling covered the fullrange of hospital activity as well asintermediate care, chronic diseasemanagement and underlyingcommunity health services. It also

Figure 5.

Source: (Hirsch, 2005)

AdverseLiving

Conditions

Safer,Healthier

Population

GeneralProtection

TargetedProtection

Upstream/Public work

Publicconcern

Professionalconcern

Effect on livingconditions

Effect oncomplications

Dying fromcomplications

Effect on incidence

Vulnerable andAfflicted Popn

Effect on vulnerability

reduction

Downstream/Professional work

PrimaryProtection

TertiaryProtection

SecondaryProtection

VulnerablePopulation

Afflicted without

Complications

Afflicted with

ComplicationsBecomingafflicted

Developingcommunications

Becomingvulnerable

Becoming no longervulnerable

Strength

Citizen Involvementand Organizing

Dependency onprofessionals

Citizen exclusionand complacency

Effect onprogression

6

leading edge 10 The potential of system dynamics

explored key links and impactsbetween sectors. It was developedto be sensitive to demographicchanges within the hospital sectorand to reflect the different profiles of activity and initial capacity or

waiting lists in the 12 specialtygroupings used in the capacityplanning process. The modellingprofiled demand and capacityrequirements over a ten-year periodto 2012/13.

One output – to determine theimpact of plans on workforce – hasbuilt on the assumptions underlyingWanless workforce modellingundertaken in 2001 and 2002 basedon four Agenda for Change levels.

Learning from delayed discharge modelling

Common sense solutions can be misleadingAdding capacity in the acute sector without rebalancingresources across the whole pathway can exacerbate thesituation and cause more delays. The implication is thatincreasing capacity in periods of excess demand simplyfills that capacity. Adjusting flow variables is less costlyand more effective.

Fines may have unintended consequencesFines can cause the post-acute sector to cut services, soexacerbating both delayed discharge and waiting lists.The effects of service cuts may spill over into otherareas. If fines are levied they need to be re-investedfrom a whole-systems perspective.

Interventions that can help include:

• re-balancing resources across all sectors, not justadding to hospital capacity

• balancing flows through different routes/reducinglengths of stay – better than capacity

• addressing variation in flows – particularly loops likere-admission rates.

Keeping people out of hospital is more effective thantrying to get them out faster.

Figure 6. Sectors of the delayed discharge model

Source: Symmetric SD

Primary careHospital

assessment

Electivewait list

Medicalbeds (slow)

Medicalbeds (fast)

Transfersmedical to

surgical

Surgical beds

Intermediatecare

Nursingresidential

Domicilliarycare

NHScontinuing

care

Routes to home exist fromeach sector

Surgicalemergencies

Medicalemergencies

Hospitalavoidance

7

Each resultant staff group was thenlinked with an appropriate ‘capacitydriver’ within the model. Data fromthe model was then re-aggregated todetermine the change in skills mix andlocation as well as to indicate directcosts to the system.

The outputs identified somesignificant challenges, particularly inthe anticipated growth in the acutesector workforce, with only modestincreases in the communityworkforce. The simulator outputs hadsuggested that more focus should begiven to developing the community,and in particular the intermediatecare, workforce.

Other applications

A number of other studies have usedsystem dynamics to investigate issuesarising from the application ofnational frameworks to support localhealth and social care systems. Thesehave centred on older people’s andmental health services, and differ fromthe specific national-level modellingin a number of ways. In particular,they have involved:

• future capacity requirements andplanning hospital services in NorthWest London SHA

• changes in depression servicedelivery

• development of a commissioningstrategy for adult learningdisabilities

• future proofing investment inbuildings

• testing strategies for reconfiguringcommunity services

• flexible use of bed-based andcommunity services across whole-health and social careeconomies

leading edge 10 The potential of system dynamics

• testing of commissioning options

• testing of major capital investmentplans for acute hospitals in the light of pre and post-acute trendsin treatment

• examination of hospital admissionprocesses, post-acute capacity andintermediate care

• investigation of intermediate careservices to assist timely dischargeand for admission avoidance

• reallocation of community bedsacross the whole patient pathwayfor older people

• forensic strategy and prisonhealthcare

• managing mental health cases inthe community, to avoidinstitutional care

• testing of guidelines for depression,with the Core ServicesImprovement Partnership (CSIP)

• disease management.

In each case the focus has been ontesting the effect of the changes onthe whole pathway as well as on theagency leading the change.Discrepancies that emerge fromthese analyses can often lead to:

• exposing the informal policieswhich an organisation uses whendelays/stocks are not permitted, forexample in A&E

• questioning the validity of dataitself – for example, whether thedata is a function of managementaction rather than an absolutereflection of the characteristics of patients

• identifying implications forinformation system design – that is,the data that might need to becollected for improved managementof the patient pathways, rather thanfor traditional performancemanagement indicators

• highlighting conflict betweenachieving sustainable outcomesand applying short-term copingactions when demand pressuresare high.

Maps and models are constructed inrelatively inexpensive purpose-builtsoftware (for example, iThink andPowersim) for desk-top computers.The key is to produce the simplestmodel possible, consistent withmaintaining its transparency andhaving confidence in its ability to cast new light on the issue ofconcern. This means keeping theresolution of the model at thehighest level, so distinguishing itfrom most spreadsheets and process maps.

Should you use systemdynamics?

Those using system dynamics haveexperienced a change in perceptionsand thinking; many are progressingtowards embedding it as a planningtool to support traditional approaches.

Of course, to perform effectively,system dynamics must both bewell-informed and used correctly. It isbest applied as a strategic tool and isnot necessarily the most appropriateapproach for operational issues whichlack repeatability and have no obviouspatterns. For it to work well requires:

• defined issues which have provedresilient to more conventionalapproaches

‘A number of studies have usedsystem dynamics to investigateissues arising from the applicationof national frameworks to support local health and socialcare systems’

8

The NHS Confederation1 Warwick Row London SW1E 5ER

Tel 020 7959 7272 Fax 020 7959 7273E-mail enquiries@nhsconfed.org

www.nhsconfed.org

The NHS Confederation29 Bressenden Place London SW1E 5DD

Tel 020 7074 3200 Fax 020 7074 3201E-mail enquiries@nhsconfed.org

www.nhsconfed.org

Further copies can be obtained from:

NHS Confederation DistributionTel 0870 444 5841 Fax 0870 444 5842E-mail publications@nhsconfed.orgor visit www.nhsconfed.org/publications

LED 01001 Registered Charity no: 1090329

leading edge 10 The potential of system dynamics

• a motivated and consistentmanagement team composed ofboth issue experts and modellers;for best implementation this teammust own both the models andtheir insights

• an openness to radical change anddata sharing

• a minimisation of defensiveattitudes and a willingness tochallenge the status quo.

Overall, system dynamics is provingto be a very versatile way of bringing whole-systems thinking to life in a rigorous, testable way. Itcan challenge and question existing practices and data, and can help design and test alternativeideas for improvement over a wide range of performancemeasures.

NHS Confederation Leading edge briefings are designed to reflect and stimulate new thinking. If you would like toshare your views or would like more information on the issues covered in this briefing, please contact Sylvia Wyatt atsylvia.wyatt@nhsconfed.org

The Future Healthcare Network

The Future Healthcare Network (FHN) is a learning network which is addressing the large gap between our currentstate of knowledge about the shape of healthcare and what will be required by 2010 and beyond. FHN is made up of organisations that are at the leading edge of thinking about future developments of health services in the UK.FHN welcomes acute trusts, PCTs and ‘whole-system’ members who are facing system change and/or major capital investment.

FHN is part of the NHS Confederation which represents the majority of NHS organisations across the UK.Membership of FHN is only available to members of the Confederation.

For further information, contact Sylvia Wyatt on 020 7074 3220 or by e-mail atfuturehealthcarenetwork@nhsconfed.org or visit the FHN website at www.nhsconfed.org/fhn

References

1 Richmond, B. 1998: iTthink user manual. ISEE Systems

2 Sterman, J. 2000: Business dynamics – systems thinking and modelling for acomplex world. Irwin McGraw-Hill, Boston

3 Hirsch, G., Homer, J., McDonnell, G. and Milstein, B: Achieving health carereform in the United States: towards a whole system understanding. Paperpresented at the 2005 System Dynamics Conference, Boston, USA

4 Wolstenholme, E F. 2003: ‘Towards the definition and use of a core set ofarchetypal structures in system dynamics’, System Dynamics Review vol 19, No. 1 (Spring 2003), 7-26

5 Wolstenholme, E F., Monk, D., Smith, G. and McKelvie, D: Using systemdynamics to influence and interpret health and social care policy in the UK. Paperpresented at the 2004 system dynamics conference, Oxford, UK

Further information

Symmetric SD: www.symmetricsd.co.uk

System Dynamics Society: www.systemdynamics.org

The Whole System Partnership: www.thewholesystem.co.uk