Getting to Zero and other Possible Dreams

Don Goldmann, M.D.Senior Vice President

Institute for Healthcare ImprovementProfessor of PediatricsHarvard Medical School

No conflicts to declare

100,000 Lives Campaign Objectives

(December 2004 – June 2006)

• Avoid 100,000 unnecessary deaths in participating hospitals

• Enroll more than 2,000 facilities

• Raise the profile of the problem - and hospitals’ proactive response

• Build a reusable national infrastructure for change

Some is not a number, soon is not a time - Berwick

100,000 Lives Campaign “Planks”

• Rapid response teams• Evidence-based care for acute myocardial

infarction • Prevention of adverse drug events (medication

reconciliation) • Prevention of central line infections (Central

Line Bundle)• Prevention of surgical site infections (correct

perioperative antibiotics at the proper time and other elements of the Surgical Infection Bundle)

• Prevention of ventilator-associated pneumonia (Ventilator Bundle)

Campaign Field Operations

FACILITIES (3000-plus)

NODES (> 55)

*Each Node Chairs 1 Network

*30 to 60 Facilities per Network

Introduction, expert support/science, ongoing

orientation, learning network development,

national environment for change

Ongoing communication

IHI and Campaign Leadership

Local recruitment and support of a smaller network

through communication/collaborative

s

Implementation (with roles for each

stakeholder in hospital and use of existing spread strategies

Measurement Strategy

• Change in aggregate hospital mortality, compared to 2004, in terms of “lives saved”– Case mix adjustment from three sources, but not yet

Hospital Standardized Mortality Ration (HSMR)

• Direct submission of monthly raw mortality data (deaths/discharges) to IHI

• Optional data at the intervention-level (e.g., ventilator pneumonia rates, process measures)

100,000 Lives Campaign Results

• Estimated 120,000 lives saved by participating hospitals through overall improvement (IHI cannot attribute change in mortality to the Campaign per se – research studies pending)

• Over 3,100 Hospitals Enrolled• Over 78% of all acute care beds

• Participation in Campaign Interventions• Rapid Response Teams: 60%• AMI Care Reliability: 77%• Medication Reconciliation: 73%• Surgical Site Infection Bundles: 72%• Ventilator Bundles: 67%• Central Venous Line Bundles: 65%• All six: 42%

100,000 Lives Campaign Results• Over 55 field offices (“nodes”) and over 130 mentor

hospitals • Strong national partner support (CDC, AHRQ, Joint

Commission, ACC/AHA, etc.)• Thousands on national calls• Large increase in web activity and downloads of

Campaign tool kits• Great media coverage (Newsweek, US News and World

Report, Wall Street Journal, NY Times)• Related campaigns forming nationally and globally

(Canada, Australia, Denmark, England) • Changes in expectations for care (“getting to zero”) in

some participating facilities (many reports of zero ventilator-associated pneumonia or catheter-related BSIs)

Success Factors

• Inspiring goal and clear deadline• Easy sign-up• Minimal reporting requirements• Straightforward interventions• Optimism, personal motivation, volunteerism• Practical direction for hospital leaders

– Demonstrated link between quality and cost

• Useful tools• Vibrant, distributed national learning network• Young, dedicated field team, logistics

5 Million Lives Campaign

• A campaign against harm (injuries/adverse events)

• Harm is defined as levels e-i using NCC MERP* Index criteria– Level e is temporary harm that required intervention– Level i is death

• Harm is counted…– Whether or not it is considered preventable– Even if present on admission to the hospital if

attributable to medical care

* National Coordinating Council for Medication Error Reporting and Prevention

How did IHI Decide on 5 Million Harms?• 37 million admissions to acute care US hospitals annually

– AHA National Hospital Survey, 2005

• 40-50 level e-i harms per 100 admissions– Chart reviews in 3 hospitals using IHI Global Trigger Tool (GTT)*

• Therefore, about 15 million harms occur per year (37 million admissions X 40 harms per 100 admissions)

• If best known results can be replicated, might avoid 3.5 million harms per year = 7 million in 2 years– 5 million seemed like a good stretch goal– We know that even perfect compliance with all of the planks will not

be enough to avoid 5 million harms

• Further validation of GTT psychometrics pending

* http://www.ihi.org/IHI/Topics/PatientSafety/SafetyGeneral/Literature/IHIGlobalTriggerToolforMeasuringAEs.htm

• Reduce Surgical Complications – Adopt “SCIP”

• Prevent Harm from High Alert Medications

• Prevent MRSA Infections• Reduce Readmissions in patients with

Congestive Heart Failure• Prevent Pressure Ulcers• Get Boards on Board

5 Million Lives Campaign Planks

Tough Questions• IHI claims that organizations need to have

leadership commitment, improvement expertise and capacity, and the ability to apply QI methods (rapid cycle PDSAs) – just for starters

• But contact with many participating hospitals suggests that such capability is not widespread

• So….are we– Encouraging brute force (“hire-a-nurse”) projects to

implement a few “planks?” Relying on charismatic champions? ….or….

– Creating fertile soil for true institutional transformation?• How good is the evidence? When is it good “good

enough” to spread?– MRSA and RRTs: more later

Prevent MRSA Infection

0

2000

4000

6000

8000

10000

12000

14000

16000

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

Year

Nu

mb

er o

f re

po

rts

No susceptibility information MSSA MRSA

Mandatory surveillance introduced April 2001

Baseline year for targets 2003/04

Mandatory enhanced surveillance October 2005

S. aureus bacteraemia: methicillin sensitivity (English NHS acute Trusts,

voluntary surveillance 1990-2006)

Provisional data

Temporal trends in MRSA bacteraemia rates, by region

Provisional data

0

0.5

1

1.5

2

2.5

3

Apr 0

1 to

Jun

01

Oct 01

to D

ec 0

1

Apr 0

2 to

Jun

02

Oct 02

to D

ec 0

2

Apr 0

3 to

Jun

03

Oct 03

to D

ec 0

3

Apr 0

4 to

Jun

04

Oct 04

to D

ec 0

4

Apr 0

5 to

Jun

05

Oct 05

to D

ec 0

5

Apr 0

6 to

Jun

06

Oct 06

to D

ec 0

6

Quarter

Est

imat

ed i

nci

den

ce r

ate

per

10,

000

bed

-d

ays

East Midlands East of England London

North East North West South East

South West West Midlands Yorkshire & the Humber

Introduction of national target

Estimated overall rate increase % per quarter

Heterogeneous regional patterns 0.5

Estimated overall rate decrease 3% per quarter

Homogeneous regional patterns

MRSA in Europe

Is this remarkable variation due to:• Transmissibility and virulence of distinct

genotypes?• Size, design, or type of hospital?• Case mix?• Practice variation?

– Compliance with known, measurable evidence based practices?

– Less tangible features, such as culture and organization of an intensive care unit?

• Are nosocomial infections an “expected” consequences of caring for very sick, complex patients, or intolerable, potentially preventable adverse events

– Vermont Oxford NICQ visits to “best of breed” NICUs

A Modest Proposal…

• Improve reliability of basic infection control procedures

• Hand hygiene• Isolation procedures• Screening cultures

Reliability Science• Health care is riddled with defects

– 40-50% compliance with hand hygiene!!??– What happens at Intel…– What happens in Bowling Green…

• From the patient’s point of view, it’s “all or nothing”

• Reliability science offers robust approaches to reducing defects and harm in health care

Component vs. Composite AdherenceContact Precautions

• COMPONENT: 80% hand hygiene, gloves on entering room

• COMPONENT: 78% gowns on entering room• COMPONENT: 65% hand hygiene after

removing gloves• COMPOSITE: 50% get all three

Reliability is failure free operation over time from the viewpoint of

the patient

Defects in outpatient

asthma care

Acute asthma attack

Admission through discharge

Defects in outpatient

care

Years/Months Days Years/Months

Defects in hospital care

Defect free care overtime from the patient’s viewpoint

Levels of Reliability

• Chaotic process: Failure in greater than 20% of opportunities

• 10-1: 80 or 90 percent success: 1 or 2 failures out of 10 opportunities (no consistent articulated process)

• 10-2: 5 failures or fewer out of 100 opportunities (process is articulated by front line)

• 10-3: 5 failures or fewer out of 1000 opportunities• 10-4: 5 failures or fewer out of 10,000

opportunities

Blood banking and anesthesiology alone achievethe higher levels of reliability in medicine

• Remember, it’s “all or nothing” – not compliance with each individual component of “best practice”

• Most institutions do fairly well with individual components of evidence-based practice, but performance drops dramatically when the standard is “all or nothing”

• We are trying to decrease the “defect rate” and to achieve a reliability of performance to the 10-2 level (at least 95% compliance with the entire package of evidence-based practice)

Reliability in Healthcare

Guidelines v. Bundles (Intervention Packages)

• Guidelines tend to be long, all-inclusive, and confusing– Many potential interventions are supported by

some evidence

• Guidelines are difficult to translate into action and often are ignored by clinicians

• What if just a few key, actionable interventions, supported by strong evidence, were culled from the guidelines?

What Is a Bundle?• A grouping of best practices with respect to a

disease process that individually improve care, but when applied together result in substantially greater improvement

• The science behind the bundle is so well established that it should be considered standard of care

• Bundle elements are dichotomous and compliance can be measured: yes/no answers

• Bundles eschew the piecemeal application of proven therapies in favor of an “all or none” approach

Central Venous Catheter Bundle• Hand hygiene before inserting a catheter or

manipulating the system and catheter site• Maximal barrier precautions for line insertion

– Hand hygiene– Non-sterile cap and mask– Sterile gown and gloves– Large sterile drape

• Antiseptic prep used for catheter insertion as per hospital protocol– 2% chlorhexidine supported by evidence (but FDA

warning for neonates)• Site selection• Timely removal

Central line-associated bloodstream infection rate in 66 ICUs, Southwestern Pennsylvania, April 2001-March 2005

Pronovost et al.,N Engl J Med; 2006;355:2725Decrease from 7.7 to 1.4 per 1000 catheter days in 103 ICUs

CDC

Imagine what would happen to the MRSA infection rate in there were nearly zero central venous

catheter infections…

A Hand Hygiene “Bundle”

• Staff knowledge• Staff competency• Alcohol and gloves available at the point of care

– Operational, full dispensers providing correct volume of rub

– At least 2 sizes of gloves

• Correct performance of hand hygiene + gloves worn for standard precautions– Concurrent monitoring and feedback– Focus on leaving the bedside– Staff accountability

Prevent MRSA Infection and Colonization

• Colonized patients comprise the reservoir for transmission (“colonization pressure”)

• High rates of MRSA colonization complicate empiric antibiotic therapy (e.g., vancomycin)

• Colonized patients have a high rate of MRSA infection– Nearly 1/3 develop infection, often after discharge

• Colonization is long-lasting, and patients can transmit MRSA to patients in other health care settings (e.g., nursing homes), as well as to family members

Five Key Interventions

• Compliance with Central Venous Catheter and Ventilator Bundles

• Hand hygiene*• Active surveillance cultures (ASCs)• Decontamination of the environment

and equipment• Contact precautions for infected and

colonized patients

* Especially before contact with the patient and after contactwith the patient and environment

What Changes Can We Make? Understanding the System

PrimaryDriversOutcome

SecondaryDrivers

ProcessChanges

Aim: Animprovedsystem

P. Driver

S. Driver 1Change 1

P. Driver

S. Driver 2

S. Driver 3

S. Driver 1

S. Driver 2

Change 2

Change 3

CauseEffectDrives

PrimaryDriversOutcome

SecondaryDrivers

ProcessChanges

AIM:A New

ME!

Calories In

Limit dailyintake

TrackCalories

CaloriesOut

Substitutelow calorie

foods

Avoidalcohol

Work out 5days

Walk toerrands

PlanMeals

Drink H2ONot Soda

drives

drives

drives

drives

drives

drives

drives

drives

What Changes Can We Make?Understanding the System for Weight Loss

“Every system is perfectly designed to achieve the results that it gets”

Outcome = Structure + Process -Donabedian

How Will We Know We Are Improving?Understanding the System for Weight Loss with Measures

PrimaryDriversOutcome

SecondaryDrivers

ProcessChanges

AIM:A New

ME!

Calories In

Limit dailyintake

TrackCalories

CaloriesOut

Substitutelow calorie

foods

Avoidalcohol

Work out 5days

Walk toerrands

PlanMeals

Drink H2ONot Soda

drives

drives

drives

drives

drives

drives

drives

drives

• Weight• BMI• Body Fat• Waist size

• Daily caloriecount

• Exercisecalorie count • Days between

workouts

• Avg drinks/week

• Runningcalorie total

• % ofopportunitiesused

• Sodas/week

• Meals off-plan/week

• Avg cal/day

Etc...

Measures let us• Monitor progress in

improving the system• Identify effective changes

What Changes Can We Make? Understanding the System for Reducing Hospital Acquired Infections

See the ‘Change Package’

How Will We Know We Are Improving? Understanding the System for Reducing Hospital Acquired Infections with Measures

Active Surveillance

• Perform active surveillance cultures (ASCs) to detect colonized patients on admission– Necessity of ASCs per se in controlling MRSA is

controversial – why are we recommending it?• “Knowledge is power” – clinical cultures miss many colonized

patients and vastly underestimate the magnitude of the problem– Added value varies by institution (Huang SS: JID 2007;195:330-8)

• ASCs on admission, followed by testing weekly and/or at discharge, is necessary to document the extent of transmission and the success of control measures

• Nose +/- perineum/axilla +/- rectum and skin lesions/broken skin• Successful programs combine ASCs with reliable implementation

of other interventions– Controversy regarding ASCs for high-risk areas (ICUs) vs.

entire hospital

Evidence for ASCs

• European experience

• Control of nosocomial MRSA outbreaks

• Mathematical models

• Observational studies from individual hospitals

• Interrupted time series study

• Cluster randomized trial

Methicillin resistance

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Antimicrobial Resistance in Antimicrobial Resistance in Staphylococcus Staphylococcus aureusaureus Blood Isolates, Denmark 1960-1995 Blood Isolates, Denmark 1960-1995

DANMAP Report, 1997.Rosdahl VT et al. Infect Control Hosp Epidemiol 1991;12:83-88.

1960 1965 1970 1975 1980 1985 1990 1995

Huang SS et al., Clin Infect Dis 2006;43:971-8

Impact of Active Surveillance in ICUs

Active Surveillance

• Perform active surveillance cultures (ASCs) to detect colonized patients on admission– Necessity of ASCs per se in controlling MRSA is

controversial – why are we recommending it?• “Knowledge is power” – clinical cultures miss many colonized

patients and vastly underestimate the magnitude of the problem– Added value varies by institution (Huang SS: JID 2007;195:330-8)

• ASCs on admission, followed by testing weekly and/or at discharge, is necessary to document the extent of transmission and the success of control measures

• Nose +/- perineum/axilla +/- rectum and skin lesions/broken skin• Successful programs combine ASCs with reliable implementation

of other interventions– Controversy regarding ASCs for high-risk areas (ICUs) vs.

entire hospital

Beware….

• Pseudomonas

• Acinetobacter

• Stenotrophomonas

• Burkholderia

• ESBL and carbapenemase-producing Gram-negative bacilli

• And many others….

Weighing the Evidence

• How much evidence is required before deciding to spread change?

• What kind of evidence is appropriate?– Randomized controlled trials

• Cluster randomized trials

– Quasi-experimental studies• Statistical process control• Time-series analysis

– Qualitative studies• Behavioral science, Sociology, Anthropology

– Mixed methods

Transition from Descriptive Theory to Normative Theory – ⇧Degree of Belief

p.6

Carlile and ChristensenPractice and MalpracticeIn Management Research

47

Pawson and Tilley: Realistic Evaluation

Pawson R, Tilley N. Realistic Evaluation. London: Sage Publications, Ltd.; 1997. 48

Pawson and Tilley

Pre-Test Treatment Post-Test

Experimental Group

O1 X O2

Control Group O1 O2

The Classic Experimental Design: “OXO”

Pawson R, Tilley N. Realistic Evaluation. London: Sage Publications, Ltd.; 1997. 49

Pawson and Tilley

Context + New Mechanism = Outcome

C + M = O

Pawson R, Tilley N. Realistic Evaluation. London: Sage Publications, Ltd.; 1997. 50

Pawson and Tilley

“Programs work (have successful ‘outcomes’) only in so far as they introduce the appropriate ideas and opportunities (‘mechanisms’) to groups in the appropriate social and cultural conditions (‘contexts’).”

Is life this simple?

X Y(If only it was this simple!)

No, it looks more like this…

X3

X2

X1

X5

X4

Y

In this model there are numerous direct effects between the independent and variables (the Xs) and the dependent

variable (Y).

Time 1

Time 3

Time 2

Dependent or outcome variable

Ind

ep

en

den

t V

ari

ab

les

In this case, there are numerous direct and indirect effects between the independent variables and the dependent variable.

For example, X1 and X4 both have direct effects on Y plus there is an indirect effect due to the interaction of X1 and X4

conjointly on Y.

Y

Or, probably more like this…

X3

X2

X1

X5

X4

Time 1

Time 3

Time 2

R3

R2

R1

R5

R4

RY

Key Reference on Causal Modeling

Blalock HM, ed. Causal Models in the Social Sciences. Chicago:

Aldine; 1999.

R = residuals or error terms representing the effects of variables

omitted in the model.

Rigorous Learning in Complex Systems

SimpleLinear

Cause-and-Effect

ComplexNon-Linear

Chaotic

“Rigorous”Learning

Poor Learning

TraditionalRCTs

Case Series“Anecdotes” Static RCTs

•“Dynamic” Cluster RCTs•Statistical Process Control•Time Series Methods•Mixed Methods•Anthropology•Ethnography•Journalism

Weighing the Evidence• How much evidence is required before

deciding to spread change?• What kind of evidence is appropriate?

– Randomized controlled trials• Cluster randomized trials

– Quasi-experimental studies• Statistical process control• Time-series analysis

– Qualitative studies• Behavioral science, Sociology, Anthropology

– Mixed methods

The Case of Rapid Response Teams

• “Early trials of medical emergency teams suggested a large potential benefit – to the point that some observers regarded further study as unethical. However, a large, randomized trial subsequently showed that medical emergency teams had no effect on patient outcomes.”

Auerbach, et al., NEJM 2007:357:608-613

The MERIT Cluster Randomized Trial

• 23 Australian hospitals randomized• 2-month baseline, 4-month preparation period, 6-

month intervention• Superb statistical analytic plan• More inter- and intra-hospital variance than expected,

much lower event rate than expected• Increased call rate in intervention hospitals, but no

effect on outcomes– Reduction in mortality in both arms of study

• Sub-optimal team activation in patients with call criteria

MERIT Study Investigators, Lancet 2005;365:2091-2097

What If….

• Baseline period was used to adjust power– Study would have been “futile”

• Performance data were fed back in real time

• QI was encouraged to improve performance

• Mixed methods were used to understand context and outcomes in individual sites

Lessons

• Every QI “experiment” should use the most appropriate evaluation method for the question and context

• The broadest possible palette of methods should be utilized

• No opportunity to learn should be wasted