Context in Improvement Science:

Rapid Scale Up of High Impact Interventions for Improved Child Survival in Ghana -Reflection from 202 Public Hospitals

Sodzi Sodzi-Tettey, MD, MPH

Director, PFA!PFHS Webinar

October 23rd 2015

Introduction

Design

The 9 Hospital Prototype

Results – Prototype - Scale

Reflections – Scale & Sustainability

350

18580

4030

0

50

100

150

200

250

0

100

200

300

400

500

600

700

800

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

U5

an

d N

eon

atal

Mo

rtal

ity/

1

,00

0 L

ive

Bir

ths

MM

R/1

,00

,00

0 L

ive

Bir

ths

Maternal MortalityRate

U5 Mortality Rate

Neonatal Mortality

Source: World Bank http://data.worldbank.org/indicator/SH.STA.MMRT (1990 figure is from UNDP http://www.undp-gha.org/mainpages.php?page=MDG%20Progress)

AIM:Assist and accelerate Ghana’s efforts to achieve

Millennium Development Goal 4 (66%

reduction in Under-5 mortality to 40/1000 live births by 2015)

through the application of quality improvement methods

Funded by the Bill & Melinda Gates Foundation

COLLABORATORS :

• Ambitious Aims• Systems View • Core Metrics with Feedback • Rapid Cycle Tests of local

ideas

DESIGNQI Team Members at a Meeting at

OLGH, Asikuma

• Multidisciplinary

What are we trying toaccomplish?

How will we know that achange is an improvement?

What change can we make thatwill result in improvement?

Model for Improvement

PlanAct

DoStudy

Source: Associates for Process Improvement

Change package of process improvements that had been shown to be effective in similar contexts

Assessment and Design

Period

Learning Session 1

© Institute for Healthcare Improvement

Learning Session 2

ACTIVITY PERIOD

Repeated improvement

cycles:

Learning Session 3

12 -24 months

Intensive support from project staff & DHMT

ACTIVITY PERIOD

Repeated improvement

cycles:

Improvement Collaborative Network

Health Facilities

Reducing Under 5 Deaths in NCHS Hospitals

Delay in Seeking Care

Reliable use of Protocols

Delay in Providing Care

We documented our theory of what leads to U5 deaths in hospitals - Driver Diagram

1o Drivers 2o Drivers

Mobilizing Community

Cultural Barriers

Referral from 1o facility

Financial Barriers

Attractiveness of services

Emergency response Syst.

Outpatient services

Staff Issues

Admission Process

Process Measures

Staff Knowledge and Skills

Availability of Drugs, supplies and equipment

Access to Protocols

Outcome

Average cervical dilatation of women in labour arriving at Hospital

Average time of 1st encounter with hospital after onset of symptoms for children U5

Average Time critically ill U5 identified in hospital to

time first treatment is commenced

Percentage adherence to selected protocols

Average stock out for antimalarial, blood and oxygen

Knowledge of 1o caregiver

Average Time spent by woman in labor from registration until assessment by midwife of doctor

Institute for Healthcare Improvement, 2013Slide 8

Start-up:

months

1 – 8

Total Pop’n:Under 5 Pop’n:

Nov 2007

Wave 1:

months

9 – 22

350,000

60,000

Jul 2008

Wave 2:

months

23 – 63

5 million500,000

Sept 2009

Wave 1R:

months

58 – 89

11 million1.7 million

Aug 2012

Start Small, Scale up Rapidly with Change Package

No of. QI Teams: 30 258 350 369>1,046

Jan 2013

Wave 3:

months

24 – 89

11 million1.7 million

Oct 2009

Wave

4:

months

63 – 89

22 million3.3 million

*Referral project launch

41 Referral Teams

Pareto Principle

Roughly 80% of the problems are caused by only 20% of the contributors

Results give credibility & enhance buy in

Is there standardized data to measure the effect of specific changes tested and implemented?

Are processes improving?

Is process improvement influencing outcome?

Are you focusing on key processes?

What adaptations in design can be made?

22.37.8

UCL

LCL0

50

100

Jan

-08

Ma

r-0

8M

ay-0

8Jul-

08

Se

p-0

8N

ov-0

8Jan

-09

Ma

r-0

9M

ay-0

9Jul-

09

Se

p-0

9N

ov-0

9Jan

-10

Ma

r-1

0M

ay-1

0Jul-

10

Se

p-1

0N

ov-1

0Jan

-11

Ma

r-1

1M

ay-1

1Jul-

11

Se

p-1

1N

ov-1

1Jan

-12

Ma

r-1

2M

ay-1

2Jul-

12

Se

p-1

2N

ov-1

2

De

ath

s p

er

10

00

A

dm

issio

ns

Under 5 Deaths per 1000 Admissions in Our Lady of

Grace Hospital, Breman Asikuma, (Jan 2008 - Dec 2012),

U-Chart

Rate

17.4

8.8

UCL

LCL0

20

40

60

Jan

-08

Ma

r-0

8M

ay-0

8Jul-

08

Se

p-0

8N

ov-0

8Jan

-09

Ma

r-0

9M

ay-0

9Jul-

09

Se

p-0

9N

ov-0

9Jan

-10

Ma

r-1

0M

ay-1

0Jul-

10

Se

p-1

0N

ov-1

0Jan

-11

Ma

r-1

1M

ay-1

1Jul-

11

Se

p-1

1N

ov-1

1Jan

-12

Ma

r-1

2M

ay-1

2Jul-

12

Se

p-1

2N

ov-1

2

Death

s p

er

1000

Ad

mis

sio

ns

Under 5 Deaths per 1000 Admissions in St. Francis Xavier Hospital, Assin Foso, (Jan 2008-

Dec 2012), U-Chart

Rate

25.817.5

UCL

LCL0

20

40

60

Jan

-08

Ma

r-0

8M

ay-0

8Jul-

08

Se

p-0

8N

ov-0

8Jan

-09

Ma

r-0

9M

ay-0

9Jul-

09

Se

p-0

9N

ov-0

9Jan

-10

Ma

r-1

0M

ay-1

0Jul-

10

Se

p-1

0N

ov-1

0Jan

-11

Ma

r-1

1M

ay-1

1Jul-

11

Se

p-1

1N

ov-1

1Jan

-12

Ma

r-1

2M

ay-1

2Jul-

12

Se

p-1

2N

ov-1

2

De

ath

s p

er

10

00

A

dm

iss

ion

s

Under 5 Deaths per 1000 Admissions in St. Martins de Porres Hospital, Eikwe, (Jan

2008 - Dec 2012), U-Chart

Rate

18.4 8.0

UCL

LCL0

20

40

60

Jan

-08

Ma

r-0

8M

ay-0

8Jul-

08

Se

p-0

8N

ov-0

8Jan

-09

Ma

r-0

9M

ay-0

9Jul-

09

Se

p-0

9N

ov-0

9Jan

-10

Ma

r-1

0M

ay-1

0Jul-

10

Se

p-1

0N

ov-1

0Jan

-11

Ma

r-1

1M

ay-1

1Jul-

11

Se

p-1

1N

ov-1

1Jan

-12

Ma

r-1

2M

ay-1

2Jul-

12

Se

p-1

2N

ov-1

2

De

ath

s p

er

10

00

A

dm

iss

ion

s

Under 5 Deaths per 1000 Admissions in Margaret

Marquart Hospital, Kpando, (Jan 2008 - Dec 2012), U-Chart

Rate

44.4

25.815.6

UCL

LCL0

20

40

60

Jan

-08

Ma

r-0

8M

ay-0

8Jul-

08

Se

p-0

8N

ov-0

8Jan

-09

Ma

r-0

9M

ay-0

9Jul-

09

Se

p-0

9N

ov-0

9Jan

-10

Ma

r-1

0M

ay-1

0Jul-

10

Se

p-1

0N

ov-1

0Jan

-11

Ma

r-1

1M

ay-1

1Jul-

11

Se

p-1

1N

ov-1

1Jan

-12

Ma

r-1

2M

ay-1

2Jul-

12

Se

p-1

2N

ov-1

2

Death

s p

er

1000

Ad

mis

sio

n

Under 5 Deaths per 1000 Amdissions in Catholic

Hospital, Battor, (Jan 2008 -Dec 2012), U-Chart

Rate

23.3

UCL

LCL0

20

40

60

1…

3…

5…

7…

9…

1…

1…

3…

5…

7…

9…

1…

1…

3…

5…

7…

9…

1…

1…

3…

5…

7…

9…

1…

1…

3…

5…

7…

9…

1…

Death

s p

er

1000

Ad

mis

sio

ns

Under 5 Deaths per 1000 Admissions in Mathias Hospital,

Yeji, (Jan 2008 - Dec 2012), U-Chart

Rate

13.5

Weak management support Poor team dynamics High Attrition of core QI team

members Challenged reporting of process

measures

Driver

Area of

Clinical/

Community

Care Change Concept

Package

# Description of Successful Change Ideas

Delay in

Seeking

Care

Care–seeking

behaviour

Targeted health

education

1A Targeted health education on early care-seeking using

interactive platforms

1B Community engagement and education via durbar or place

of worship

Referral

Engaging

primary

providers

1C Engagement with health providers (both traditional and

allopathic)

Prompt

Diagnosis and

Treatment

Triage

2A

Triage system for screening and emergency treatment of

critically ill children

Separate U5 OPD services from adult OPD service

Prioritize U5 outpatient care

Prioritize U5 inpatient care

Delay in

Providing

CareFast Track

Non-

Adherence

to

ProtocolsAdherence to

Protocols

Training/

Coaching/

Mentoring

3A

Training staff on protocols followed by regular coaching and

mentoring which include ad hoc testing on site with

immediate feedback.

3B

Training postpartum women and other care givers on

hygienic cord care through demonstration, practice and

immediate feedback. Midwives and nurses teach,

3C

Mother-to-mother support group on food choices and

frequency of feeding while on admission under mentoring of

nurses.

Task-shifting 3D Empowering nurses to start acting on standard treatment

protocols before doctor arrives

Hospital Change Package

202H

68H

32H

9H

Drivers of Hospital Based Deaths

% of QI Teams Adopting at least one Change Idea (N=134)

Comments

Early Care Seeking 84.3 Three Change Ideas (H-1A, 1B, 1C)

Prompt Provision of Care 69.4 A Change bundle (H-2A)

Adherence to treatment protocols

69.4 Four Change Ideas (H-3A to 3D)

Change Idea H -1A H-1B

H-1C

H-2A

H-3A

H-3B

H-3C

H-3D

Proportion of teams testing this change Idea

58.2 23.1 3.0 69.4 43.3 1.5 2.2 22.4

15.9

10.9

0

5

10

15

20

25

Ja

n-1

2M

ar-

12

Ma

y-1

2Ju

l-1

2S

ep

-12

No

v-1

2Ja

n-1

3M

ar-

13

Ma

y-1

3Ju

l-1

3S

ep

-13

No

v-1

3Ja

n-1

4M

ar-

14

Ma

y-1

4Ju

l-1

4S

ep

-14

No

v-1

4

COLLABORATIVE (134 HOSPITALS) IN 7 REGION

Subgroup Center

7.7

5.0

0

2

4

6

8

10

12

14

Ja

n-1

2M

ar-

12

Ma

y-1

2Ju

l-1

2S

ep

-12

No

v-1

2Ja

n-1

3M

ar-

13

Ma

y-1

3Ju

l-1

3S

ep

-13

No

v-1

3Ja

n-1

4M

ar-

14

Ma

y-1

4Ju

l-1

4S

ep

-14

No

v-1

4

Collaborative (135 HOSPITALS IN 7 REGION

Subgroup Center

Pressure to start at scale

Sustainability

To superimpose new learning on old system or to redesign old system with new learning?

10 Regional Quality

Advisors

~ 3000 Site Visits

~ 4000 frontline workers

trained in LSs

July 2008

Wave 1 Launch.

Nov.12 PFA end

July 2010

NCE-1

(PNC

Policy)

Mar. 2014. Cost Extension. National Scale Up. August 2015 PFA! end2012 End of

Project - initial

May 2011. Referral Supplemental. May 2015, PFA! end

Nov. 2014. NCE2. Dec. 2015 PFA! end

• Tested Feasibility of PNC Policy• PFA!- Adapted National DQI Protocols Codesigned Referral Registers & Forms Adapted Community-Facility

Collaboratives Assisted to Accelerate Mortality

Reduction ~ 400 Improvement Coaches

Clinical Skills/Jhpieg

oDHIM

S

QI Projects 1. Ghana

Systems for Health

2. UNICEF’s Newborn Project

Adaptive designing may be the exception rather than the norm – decide early whether you want to work in a real health system or carry out an experiment

Attribution is the elephant in the room –extremely important but rarely openly acknowledged by partners and stakeholders

QI alone is not enough; Quality Planning, Quality Control & Quality Improvement integration is key

Rapid tests of change help to determine what works quickly in a cost-effective manner

Empowered teams adapting contextually-relevant changes can achieve rapid impact at scale

It takes strategic alignment with health system structures/ priorities and synergy between management and frontline workers to achieve impact at scale