+ All Categories
Home > Documents > Robert L Wears, MD, MS University of [email protected]@ufl.edu Imperial College London...

Robert L Wears, MD, MS University of [email protected]@ufl.edu Imperial College London...

Date post: 27-Dec-2015
Category:
Upload: lorin-wilcox
View: 212 times
Download: 0 times
Share this document with a friend
51
Robert L Wears, MD, MS University of Florida [email protected] Imperial College London École des Mines de Paris Why Has Crowding Been Intractable? Views from System Dynamics and Resilience Engineering ED / Hospital Overcrowding “There are no side effects. There are only effects.” J Sterman
Transcript

Robert L Wears, MD, MS

University of [email protected] College LondonÉcole des Mines de Paris

Why Has Crowding Been Intractable?Views from System Dynamics and Resilience Engineering

ED / Hospital Overcrowding

“There are no side effects. There are only effects.” J Sterman

ED / Hospital Overcrowding

Overview and goals

Introduction to system dynamics methodsCausal loop diagrams, stocks & flows, dynamic simulations

System dynamics and overcrowdingPrototypical causal loops

Modeling ED / hospital overcrowding

Potential implications from a SD approachUnderstanding – how did the present come about?

Action – what can we do, what should we not do?

ED / Hospital Overcrowding

4 years ago …

ED / Hospital Overcrowding

James Fallows’ question

How is it that a system that is –so technologically advanced and

operated by such smart people

who are all working very hard –

performs so poorly?

ED / Hospital Overcrowding

James Fallows’ question

How is it that a system that is --so technologically advanced and

operated by such smart people

who are all working very hard –

performs so poorly?

Is this as good as it gets?

ED / Hospital Overcrowding

Is there something in the structure of the system?

Work patterns, peak & valley variation, artefacts, organisational policies, goals, etc all play a role

But are they sufficient explanations? Is there more?

Are there factors that can explain overcrowding, and consequent resilient or brittle behaviours of the system at some higher level of abstraction?

ED / Hospital Overcrowding

System dynamics models

Developed in ’60s in control engineering (Maruyama)

Popularized in 70-80s (Forrester, Sterman)

Used mostly in business settings (unfortunately?)

Useful in:Explaining counter-intuitive phenomena, especially in complex

sociotechnical systems, when effects are time-delayed, multiple feedback loops, etc

Determining where (where not) to intervene

ED / Hospital Overcrowding

Fundamental lessons from system dynamics

System structure influences system behaviour“Systems cause their own crises, not external forces or individuals’

mistakes”

Structure in systems is subtle“Structure” = basic interrelationships among variables that control

behaviour

“Policy resistance”, “unintended consequences”, “intractability” come from lack of system thinking

“Yesterday’s solution becomes today’s problem”

Crowding a problem since mid 1980sQuarter century of work

No progress, in fact, worse

ED / Hospital Overcrowding

System dynamic methods

Causal loop diagramsFeedback, positive & negative

Delays

Dynamic simulationsStocks

Flows

ED / Hospital Overcrowding

Causal loop diagrams

Positive word ofmouth

ED visits

Satisfied patients

+

+

+

ED / Hospital Overcrowding

Causal loop diagrams

Reinforcing loopPositive feedback

Positive word ofmouth

ED visits

Satisfied patients

+

+

+

ED / Hospital Overcrowding

Reinforcing loops

Growth, often exponentialBandwagon effect, compound interest, bacterial growth, …Rate of change increases

Virtuous cycleGood service → more business → more money → better people,

equipment → more good service …Exercise → sense of well-being → more exercise

Vicious cyclePerceived gasoline shortage → “topping off” → lines at stations →

greater perceived shortage …

Often unstableRun on bankArms raceGas crises

ED / Hospital Overcrowding

Reinforcing loop behaviour – exponential growth

0

10

20

30

40

50

60

0 5 10 15 20 25

Time

ED / Hospital Overcrowding

Causal loop diagrams

Corrective actions

ED throughput

Discrepancy

+

-

+

Desired EDthroughput

+

ED / Hospital Overcrowding

Causal loop diagrams

Balancing loopNegative feedback

Corrective actions

ED throughput

Discrepancy

+

-

+

Desired EDthroughput

+

ED / Hospital Overcrowding

Balancing loops

Goal-seeking behaviourHomeostatic

Stabilizing

Rate of change decreases

Thermostat, physiologic autoregulation, radioactive decay

Responsive to change in goal state

Resist all other changes

ED / Hospital Overcrowding

Balancing loop behaviour – goal seeking

0

2

4

6

8

10

12

0 5 10 15 20 25 30

Time

ED / Hospital Overcrowding

Causal loop diagrams – delays

Corrective actions

Current watertemperature

Discrepancy

+

-

+

Desired watertemperature

+

Delay

ED / Hospital Overcrowding

Balancing loop & delay behaviour – damped oscillation

0

2

4

6

8

10

12

0 5 10 15 20 25 30

Time

ED / Hospital Overcrowding

Balancing loop with delays

Oscillation and goal-seekingBalance depends on competing magnitudes of action and delay

More vigorous action → greater instability

Make haste slowly!

ED / Hospital Overcrowding

Delays are common

Corrective actions

ED throughput

Discrepancy

+

-

+

Desired EDthroughput

+

DelayDelay Delay

Delay

measurement,reporting, perception

administrative,decision-making

actions don't haveimmediate effects

ED / Hospital Overcrowding

Stocks and flows

“ED visits” in previous examples is a composite

Components are:Rate of new arrivals (eg, pts per hour)

Number of pts currently in ED

Rate of departures

ED / Hospital Overcrowding

Input – throughput – output model

Pts pendingArrival rate Disposition rate

Desireddisposition time

Desireddisposition rate

Work capacity

+

-

++

-

ED / Hospital Overcrowding

Basic elements combine to represent complex systems

No limit to the possible ways to combine reinforcing, balancing loops, delays, stocks, flows

Some archetypical forms occur over and overExponential growth

Goal seeking

Oscillation

Complex, nonlinear interactionsSigmoid shaped growth

Sigmoid growth w/ overshoot, oscillation

Growth and collapse*

Random

Chaotic

ED / Hospital Overcrowding

Growth & collapse – a cautionary tale?

EM statusGrowth rateConsumption of

resources

Increase ingrowth rate

Resourceadequacy

Carrying capacity

+ +

+

-

+

+

-

+

ED / Hospital Overcrowding

A cautionary tale

0

20

40

60

80

100

120

140

160

0 5 10 15 20 25 30 35

Time

0

0.2

0.4

0.6

0.8

1

1.2

ED / Hospital Overcrowding

Hysteresis

0

20

40

60

80

100

120

140

160

0 5 10 15 20 25 30 35

Time

0

0.2

0.4

0.6

0.8

1

1.2

ED / Hospital Overcrowding

Predator-prey example

ED / Hospital Overcrowding

Predator-prey example

ED / Hospital Overcrowding

The ‘balance of nature’?

3,000 Rabbit200 Fox

2,250 Rabbit150 Fox

1,500 Rabbit100 Fox

750 Rabbit50 Fox

0 Rabbit0 Fox

0 5 10 15 20 25 30 35 40 45 50Time (Year)

ED / Hospital Overcrowding

The ‘balance of nature’?

3,000 Rabbit200 Fox

2,250 Rabbit150 Fox

1,500 Rabbit100 Fox

750 Rabbit50 Fox

0 Rabbit0 Fox

0 5 10 15 20 25 30 35 40 45 50Time (Year)

ED / Hospital Overcrowding

Response to an insult

What will happen to foxes if drought cuts rabbit population in years 16 – 17?

3,000 Rabbit200 Fox

2,250 Rabbit150 Fox

1,500 Rabbit100 Fox

750 Rabbit50 Fox

0 Rabbit0 Fox

0 5 10 15 20 25 30 35 40 45 50Time (Year)

ED / Hospital Overcrowding

Almost nothing

Rabbit and Fox Populations

3,000 Rabbit200 Fox

1,500 Rabbit100 Fox

0 Rabbit0 Fox

0 5 10 15 20 25 30 35 40 45 50Time (Year)

Rabbit Population : Current RabbitFox Population : Current Fox

ED / Hospital Overcrowding

Objectives

To use system dynamic modeling as a way to illuminate resilience (or collapse) related to ED overcrowding

To identify more general underlying models of resilience / collapse in complex sociotechnical systems

To identify factors associated with performance that could inform organisational policy / procedure

ED / Hospital Overcrowding

Modeling process

Two levels of modelingScope hospital (not ED) level

Societal (emergency medicine) level

Two-pronged approachAbstract models displaying interesting behaviours

Calibrated models expressive of domain constituencies in multiple sites

Iteration between these two modes

ED / Hospital Overcrowding

Simplest model: input – throughput – output

ED / Hospital Overcrowding

Response to challenge

15 Pts/Hour60 Pts

7.5 Pts/Hour30 Pts

0 Pts/Hour0 Pts

0 12 24 36 48 60 72 84 96Time (Hour)

Arrival Rate : Current Pts/HourDeparture Rate : Current Pts/HourWorkload : Current Pts

ED / Hospital Overcrowding

Catastrophic dynamics

15 Pts/Hour60 Pts

7.5 Pts/Hour30 Pts

0 Pts/Hour0 Pts

0 12 24 36 48 60 72 84 96Time (Hour)

Arrival Rate : Current Pts/HourDeparture Rate : Current Pts/HourWorkload : Current Pts

ED / Hospital Overcrowding

Simple model, augmented w/ adaptive capacity

ED / Hospital Overcrowding

Adaptive dynamics

ED / Hospital Overcrowding

Effect of memory of adaptations

ED / Hospital Overcrowding

Simple model conclusions

Highly simplified, input-throughput-output model can demonstrate brittleness, resilience, and adaptation

But:It’s a tautology

And domain experts won’t buy it, it’s too simple

ED / Hospital Overcrowding

Feedback from domain

Input – throughput – output far too simpleToo much is packed into ‘output’

Multiple compartment models

Multiple additional effects?Cyclical?

Acuity?

Temporal – arousement, fatigue, etc

ED / Hospital Overcrowding

Extended models

ED workload Hosp workloadED admitED arrival

ED discharge

Hosp discharge

Elective admit

ED capacity Hosp capacity

ED process time

Hosp process time

ED congestion Hosp congestion

ED cycle time Hosp cycle time

ED dischargefraction

EDdischarge

home

ED pt recover

ED bounceback

ED bouncefraction

ED bounce time

Boarding pts

Ward Transfer

Boarding time

Transfer processtime

Eff HC onboarding time

Total ED census

Fraction boarding

Admitting officebypass

Hosp DCbouncebacks

FP admits

ED / Hospital Overcrowding

More interesting observations

Resilience from what point of view?

Ironies of process improvement

Co-dependency

ED / Hospital Overcrowding

Disjoint views of resilience

400

300

200

100

0

0 250 500 750 1000 1250 1500 1750 2000 2250 2500Time (Hr)

Pts

ED workload : ED capED workload : ED cap x 2ED workload : ED cap x half

ED / Hospital Overcrowding

Disjoint views of resilience

600

450

300

150

0

0 250 500 750 1000 1250 1500 1750 2000 2250 2500Time (Hr)

Pts

Hosp workload : ED capHosp workload : ED cap x 2Hosp workload : ED cap x half

ED / Hospital Overcrowding

Addiction, co-dependency

Insufficient inptbeds

Hold admits inhallway

Improve inpt flow /ED capacity

Discretionaryenergy

+

-

+

-

-

-

DelayDelay

ED / Hospital Overcrowding

Irony of improvement

0

50

100

150

200

250

300

350

0 1000 2000 3000 4000 5000 6000

ED load

ED

th

rou

gh

pu

t

ED / Hospital Overcrowding

Summary & Conclusions

“All models are wrong, but some models are useful”-- George E P Box

Very simple models can demonstrate resilient / brittle behaviours

Simple models can suggest:Complex mixture of gains and losses 2° to crowding

Perverse effects of improvement attempts

Origins of intra-organisational conflict

Building / restoring ‘capacity’ may be more useful than limiting volume

To be continued …

ED / Hospital Overcrowding


Recommended