EPIC
Evidence-based Practice
Identification and Change
Past, Present, and Future
Shoo K. Lee, MBBS, FRCPC, PhD
Director, Canadian Neonatal Network™
Scientific Director, iCARE
Professor of Pediatrics, University of Alberta
EPIC/PHSI Training Workshop
November 9 & 10, 2006
Toronto ON
Presentation Objectives
• Overview how EPIC evolved
• Describe the science behind EPIC
• Describe future EPIC plans
Background
• Continuous Quality Improvement (CQI)
methods have been investigated for
reducing bronchopulmonary dysplasia
(BPD) and nosocomial infection (NI) in the
NICU
• Limitation - existing CQI techniques employ
a subjective, uncritical approach to practice
change that may not be evidence based
How did EPIC Evolve?
 Problems with traditional continuous quality
improvement (CQI) approaches
 Subjective
 Not always evidence-based
 Seldom use data from institutions in question
 Mostly intra-institutional in nature
 Results are not always generalizeable
 We developed EPIC to improve upon traditional
CQI approaches
EPIC Objectives
• To develop a new scientific method for QI – EPIC
that is:
(a) Evidence-based – uses published evidence
(b) Objective – uses data from individual hospitals to
identify practices for targeted intervention
(c) Collaborative – uses a national network to share
expertise and experience
• To test whether EPIC reduces BPD and NI in a cluster
randomized controlled trial of Canadian NICUs
The Thee Pillars of EPIC
1. Objective
 Systematic reviews of evidence
2. Quantitative analysis
 Multi-centre outcomes and practices
 Identifies practices associated with outcome
variation that can be targeted for intervention
3. Utilizes collective multi-disciplinary expertise
 Infection control, quality improvement, etc
Method
•
•
•
•
•
•
•
•
Prospective cluster randomized controlled trial 12 NICUs
Randomization – 6 BPD, 6 NI
Each group Control for other
Additional controls - 5 other NICUs in CNN that were not
participating in the study
All infants < 32 weeks gestation were enrolled
Definition: (a) BPD – O2 need at 36 weeks GA
(b) NI – Positive Blood, CSF or Urine culture
2 phases (a) Baseline period (1 year)
(b) Intervention period (2 years)
Funded by Canadian Institutes of Health Research
EPIC - Baseline Period (Year 1)
•
•
•
•
•
•
•
Baseline data collection on outcomes and practices
Train multi-disciplinary hospital teams
Review of published literature
Meeting to share findings
Identify Critical Care Pathways
Qualitative research – identify barriers to change
Data analysis – identify practice differences associated
with outcome variation for targeted intervention
Data Analysis to Identify Practices for
Targeted Intervention
• Grouped Data Analysis
- compare outcome variations among NICUs
- identify non-therapy and therapy related risk factors - estimate the
attributable risk of risk factors
• Individual Hospital Data Analysis
- calculate hospital specific incidence rates
- identify hospital specific risk factors for targeted intervention
- conduct trend analysis using control charts
• Generalized linear mixed effects model
- to adjust results for the cluster randomized design
• Monte Carlo Bootstrap Simulation
- to estimate the 95% confidence limits for control charts
Therapy Related Risk Factor for NI - PICC
• Therapy related risks
- central lines,
- mechanical ventilation,
- parenteral nutrition,
- lack of enteral feeding
• 40% of nosocomial infection associated with
central lines
• PICC lines carried highest risk
Adjusted probability for developing
nosocomial infection for PICC lines
Line type
Risk-Ratio for NI
Umbilical catheters
2.0
Broviac cathethers
3.1
PICC catheters
3.5
EPIC – Intervention Period (2 Years)
•
•
•
•
•
•
•
Develop practice change strategies
Prepare supporting materials
NICU staff communication and training
Implement practice change strategies
Quarterly change cycles
Control Chart feedback
Revise strategies, reinforce change
Results
Group C
EPIC
Non-EPIC
5 NICU
12 NICU
Excluded
1 NICU
Group A
Group B
NI
BPD
NI
BPD
5 NICU
6 NICU
N = 2666
N = 3275
Control
5 NICU
N = 1129
Selected Patient Characteristics
Characteristics
Number
NI
2336
BPD
2316
Control
1129
Mean Gestation (wk)
Mean Birthweight (kg)
Mean SNAP-II
Male sex (%)
28.5
1246
11.2
57.2
28.9
1315
9.8
55.5
28.9
1150
12.6
56.3
Outborn (%)
Cesarean section (%)
Apgar <7 at 5 min (%)
37.5
54.1
20.3
18.0
55.5
19.1
14.9
58.8
44.0
Antenatal steroids (%)
71.1
70.7
90.5
Group A (NI) – Incidence of NI
Theoretical Infection Rate of Infant in NIT Group, NI Study (Monte Carlo Bootstrap,
n=1000)
27.0%
25.0%
24.1%
Percentage of infected baby
23.0%
21.0%
19.5%
19.0%
18.3%
17.1%
17.0%
16.1%
15.9%
16.0%
16.5%
15.0%
13.0%
12.6%
11.0%
9.0%
Baseline
Oct Dec.2003
JanM ar.2004
AprJun.2004
JulSept .2004
Study period
Oct Dec.2004
JanM ar.2005
AprJun.2005
JulSept .2005
Group A (NI) – Incidence of
BPD
Observed Incidence Rate of CLD Baby at Week 36 in NIT Group, CLD Study
48.0%
44.0%
40.0%
Percentage of CLD baby
38.9%
36.0%
32.0%
31.7%
29.3%
28.0%
28.2%
27.3%
25.4%
24.0%
23.8%
24.2%
23.1%
20.0%
16.0%
Baseline
Oct-Dec.2003
Jan-Mar.2004
Apr-Jun.2004
Jul-Sept.2004
Study period
Oct-Dec.2004
Jan-Mar.2005
Apr-Jun.2005
Jul-Sept.2005
Group A (NI) – Duration of
Oxygen Need
Observed Length of Oxygen Support for CLD Baby in NIT Group, CLD Study
12
Mean length of oxygen support (day)
11
10
9.2
9
8.91
8.62
8.39
8
7.46
7.34
7
6.35
6.13
6
6.25
5
4
Baseline
Oct-Dec.2003
Jan-Mar.2004
Apr-Jun.2004
Jul-Sept.2004
Study period
Oct-Dec.2004
Jan-Mar.2005
Apr-Jun.2005
Jul-Sept.2005
Group B (BPD) – Incidence of
BPD
Group B (BPD) – Duration of
Oxygen Need
Group B (BPD) – Incidence of
NI
Theoretical Infection Rate of Infant in CLD Group, NI Study (Monte Carlo Bootstrap,
n=1000)
23.0%
21.0%
Percentage of infected baby
19.9%
19.0%
17.3%
17.0%
16.5%
15.0%
13.8%
13.7%
13.0%
12.7%
12.6%
11.2%
11.0%
10.4%
9.0%
7.0%
Baseline
Oct Dec.2003
JanM ar.2004
AprJun.2004
JulSept .2004
Study period
Oct Dec.2004
JanM ar.2005
AprJun.2005
JulSept .2005
Group C (Controls) – Incidence
of BPD
Observed Incidence Rate of CLD Baby at Week 36 in Control Group, CLD Study
36.0%
Percentage of CLD baby
31.0%
26.0%
24.7%
22.3%
22.6%
23.2%
22.5%
21.0%
18.6%
18.5%
16.0%
14.9%
14.7%
11.0%
6.0%
Baseline
Oct-Dec.2003
Jan-Mar.2004
Apr-Jun.2004
Jul-Sept.2004
Study period
Oct-Dec.2004
Jan-Mar.2005
Apr-Jun.2005
Jul-Sept.2005
Group C (Controls) – Duration of
Oxygen Need
Observed Length of Oxygen Support for CLD Baby in Control Group, CLD Study
22
Mean length of oxygen support 9day)
20
18
16
14.64
14
13.73
13.33
12.76
12.58
12
11.88
11.48
10.43
10
8.62
8
6
4
Baseline
Oct-Dec.2003
Jan-Mar.2004
Apr-Jun.2004
Jul-Sept.2004
Study period
Oct-Dec.2004
Jan-Mar.2005
Apr-Jun.2005
Jul-Sept.2005
Group C (Controls) – Incidence
of NI
Percentage of NI(ever infected)
control group (Non-EPIC group)
25.0%
20.0%
15.0%
14.5%
12.3%
16.0%
14.0%
13.7%
11.8%
10.0%
5.0%
10.3%
7.2%
6.0%
0.0%
baseline Oct03- Jan04Dec03 Mar04
Apr04- Jul04- Oct04- Jan05- Apr05- Jul05Jun04 Sept04 Dec04 Mar05 Jun05 Sept05
Quarter
Mortality, ROP, IVH
Group A (NI)
Group B (BPD)
Group C (Control)
Baseline
8th
quarter
P
value
Baseline
8th
quarter
P
value
Baseline
8th
quarter
P
value
5.7
5.4
NS
5.0
4.2
NS
6.0
3.3
NS
ROP
>stage 3
9.5
8.5
NS
4.8
5.4
NS
5.1
7.9
NS
IVH
>grade 3
10.3
9.9
NS
7.8
9.6
NS
8.5
14.6
NS
Mortality
Conclusions
• EPIC is effective at reducing NI and BPD
in the NICU
• Interventions targeting one outcome may
affect other outcomes
• EPIC may be more effective and less
costly at improving quality of care than
traditional CQI methods
EPIC Research Program
EPIC Process
EPIC-I
2002-2005
EPIC/PHSI
2006-2009
EPIC-II
2007-2010
NIT reduced by 60%
CLD reduced by 40%
Target multiple outcomes
Target two outcomes
Test generalizeability
Improve upon EPIC-I
Target multiple outcomes
Test generalizeability
Improve upon EPIC/PHSI
Improvements in EPIC/PHSI
 Eliminate feedback delays
 one button reports
 short term feedback & unverified data
 Decrease onus of data collection
 Use only relevant CNN data
 Facilitate communication
 Knowledge Broker
 Divide NICUs into 4 groups for quarterly
teleconferences, site visits, mentorship
 Ease implementation
 4 groups will have mix of experienced EPIC sites
EPIC/PHSI Plan
 Make what we learned in EPIC-I available to all
Canadian NICUs in EPIC/PHSI
 Training of Infection Teams – MD, RN, QI
 Introduce the EPIC interventions-best practice
template
 Review EPIC-I literature reviews
 Review qualitative findings from EPIC-I
 Barriers and facilitators to change
 Develop change strategies for each NICU
 Implementation of EPIC interventions
Acknowledgements to CIHR, Micheal Smith Foundation, &
Canadian Neonatal NetworkTM EPIC Investigators
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Khalid Aziz, Memorial U
Ross Baker, U of Toronto
Keith Barrington, McGill U
Catherine Cronin, U Manitoba
Jill Hoube, UBC
Andrew James, U Toronto
Joanne Langley, Dalhousie
David SC Lee, UWO
Shoo K Lee, U Alberta
Robert Liston, UBC
Ying MacNab, UBC
Claudio Martin, UWO
Derek Matthew, Victoria Gen H
Jochen Moehr, U Victoria
•
•
•
•
•
•
•
•
•
•
•
•
Arne Ohlsson, U Toronto
Abraham Peliowski, U Alberta
Robert Platt, McGill U
K. Sankaran, U Saskatchewan
Mary Seshia, U Manitoba
Nalini Singhal, U Calgary
Bonnie Stevens, U Toronto
Anne Synnes, UBC
Paul Thiesen, BC Children’s H
Peter Von Dadelszen, UBC
Robin Walker, U Ottawa
Elizabeth Whynot, BC
Women’s
• Robin Whyte, Dalhousie U
• John Zupancic, Harvard U