Quality Improvement Science and Patient Safety Research

1
Quality Improvement Science and Patient Safety
Research

• Dan France, Ph.D., MPH
• Center for Clinical Improvement
• Vanderbilt University Medical Center

2
Outline

• Quality Improvement
• Need for the engineering mentality/systems
  thinking in healthcare
• Patient Safety
• Student Project

3
Engineering

• Design/Analysis
• Systems Engineering
• Engineering Management
• Quality Management
• Quality Engineer
• Industrial Engineer
• Health System Engineer

4
Part I. Quality Improvement
5
Background

• Institute of Medicine (IOM) reports
• Nov 1999 To Err is Human
• March 2001 Crossing the Quality Chasm
• Brief History of QI
• Scientific Management (Taylor, 1911)
• Assembly lines
• Statistical Process Control (Shewhart, 1931)
• Quality Improvement (Deming, 1955)
• Lean Production (Womack, 1990)
• Mass customization

6
Improvement in Healthcare
System Thinking Statistical Variation Scientific
Method Psychology of Change
Expert knowledge Content knowledge
Traditional Improvement
Continuous Quality Improvement
Paul Batalden MD
7
What is Quality

• Quality is the degree to which we meet or exceed
  customer expectations

8
Quality Assurance versus Quality Improvement

• Quality Assurance
• meet a specification or standard
• Take sample measurements to measure performance
• Quality Improvement
• continual process to improve current performance
• Continual measurement and data feedback

9
The Relation Between Quality Inspection,
Regulation, Management, and Improvement
Design
Management Improvement
Redesign
Number of Providers
Research Development
Sanctions
0
Level of Quality
Inspection Regulation for Public Safety
10
IOM Definition of Quality

• Six Dimensions of Quality in Healthcare
• Safe
• Effective
• Timely
• Patient centered
• Efficient
• Equitable

11
QI is a ScienceDefined Methodology

• Focus on systems (Systems theory)
• Develop ideas for change and test them
  (Scientific method)
• Understand the variation of data measured
  continuously over time (SPC)
• Understand reasons and motivation of people to
  act on data (Common cause, special cause
  variation, diffusion of innovation)
• Use a balanced set of measures (Value compass)

12
QI is a Discipline

• QI research is funded by AHRQ and NIH
• QI research is published in peer review journals
  such as NEJM and JAMA
• QI science is taught in schools of public health,
  business schools, graduate programs in
  engineering, management and education, medical
  schools in health services research,
  biostatistics, public health
• There is a national Quality Scholars program in
  healthcare

13
Variation in PracticeInstitute of Medicine

• Overuse (eg. Antibiotics, C-Section)
• Underuse (eg. Mammography, Beta-Blockers)
• Misuse (eg. Medical errors)
• The issue is unnecessary variation
• i.e., appropriateness of care

14
Six Sigma

• Domestic Airline Fatality
• 6?
• 99.99966 Right
• Mammography Screening
• 1.7 ?
• 56 Right

15
QI is a Science Statistical ApproachVariation
and Improvement

• Lessons about Variation
• Once we begin to measure important quality
  characteristics and outcomes, we notice
  variation.
• We question measurements that display no
  variation.
• Often, single data points alone are
  uninformative, but data displayed over time can
  provide information for improvement.
• The primary purpose of understanding variation is
  to enable prediction.
• Interaction among process variables produces
  sources of variation materials, methods,
  procedures, people, equipment, information,
  measurement, and environment.

16

• A process

... a series of linked steps, often but not
necessarily sequential, designed to ...

• cause some set of outcomes to occur
• transform inputs into outputs
• generate useful information
• add value

• Walter Shewhart a system of causes

17

• Constant (convergent) systems

• follow the laws of mathematical probability

How the process behaved in the past predicts how
it should behave in the future

• non-constant (divergent) systems follow the laws
  of chaos theory

How the process behaved in the past does not
predict how it should behave in the future
18

• Random variation

• represents the sum of many small variations,
  arising from real but small causes that are
  inherent inand part ofany real process

• follows the laws of probability behaves
  statistically as a random probability function

• because random variation represents the sum of
  many small causes, it cannot be traced back to a
  root cause

• represents " appropriate " variation

19

• Assignable variation

• represents variation arising from a single cause
  that is not part of the process (system of causes)

• therefore can be traced, identified, and
  eliminated (or implemented)

• represents " inappropriate " variation

20
Registration Times

• These are actual times it took triage level 2
  patients to register in the Emergency Department
  of a hospital
• 15 67 4 14 10
• 12 54 3 7 11
• 14 83 54 17 20
• 10 53

21

• Parametric frequency distribution

Number of times observed
(Number, rate, percentage, proportion)
Value observed
22

• Parameters mean and variance

center (mean, median)
Number of times observed
(Number, rate, percentage, proportion)
spread (variance, standard
deviation, range)
Value observed
23

• Probability-based boundaries

Frequency Distribution
99
Number of times observed
(Number, rate, percentage, proportion)
0.5
0.5
2.575 std. devs.
2.575 std. devs.
Value observed
24

• Statistical Process Control Chart

Observed value
Time
25

• Random variation

Process Control Chart
(How the process behaves over time)
Observed value
T1
T2
T3
T4
T5
T6
T7
T8
T9
Time
26

• Assignable variation

Process Control Chart
(How the process behaves over time)
Observed value
T1
T9
T2
T3
T4
T5
T6
T7
T8
Time
27

• Managing assignable variation

• Find a data point that probably represents
  assignable variation (usually a statistical
  outlier)

• track it to root causes

• eliminate (or implement) the assignable cause

(React to individual fluctuations in the data)
28
(No Transcript)
29

• Tampering

Using assignable methods in an attempt to
manage random variation
Shewhart proved that tampering does not just
waste time and effort -- it seriously harms
process performance
30

• Statistical process control charts

• Show the probability that an observation arose
  from the underlying process that is,

• the probability that a particular point's
  deviation from the center represents only
  "random" variation arising from the system of
  causes that make up the process, as opposed to
  "assignable" variation representing an
  identifiable, intruding cause.

• They

• separate random from assignable variation
• based on statistical probability
• using control limits, runs, trends, and other
  patterns in longitudinal data.

31

• A trend

Psych Inpatient Admits / Month
patients
UCL 76.56000
55.000000
LCL 33.44000
patients
UCL 27.77571
8.500000
LCL 0.00000
32
QI is a Science Statistical ApproachOverall
Improvement Strategy
Process change
Remove special causes
Process change
Outcome
Stable process Common cause variation is
high Average is too high
Stable process Common cause variation
reduced Average too high
Stable process Common cause variation low Average
reduced
Unstable process Special causes present Average
is too high
33

• CAP protocol compliance

Implementation Group -- Loose Abx Compliance
Baseline
Implementation
0.8
0.7
0.6
0.5
Proportion compliant
0.4
0.3
0.2
0.1
0
-23
-21
-19
-17
-15
-13
-11
-9
-7
-5
-3
-1
1
3
5
7
9
11
13
15
17
Month relative to CPM implementation
P chart - 0.01 control limits
34

• Using data to improve

• The minimum standard an annotated time series

Start with a run chart (80 of total value)
1.
Add center and goal lines (anchors the eye - now
95 of total value)
2.
Add control limits (in appropriate zones)
3.
35

• "Teen use turns upward"

high school seniors who smoke daily
1992 17.3 1993 19.0
USA Today, June 21, 1994
36

• "Teen use turns upward"

high school seniors who smoke daily
1984 18.8 1985 19.6 1986 18.7 1987
18.6 1988 18.1 1989 18.9 1990 19.2 19
91 18.2 1992 17.3 1993 19.0
(average moving range 0.778)
USA Today, June 21, 1994
37

• high school seniors smoking

38

• high school seniors smoking

Mean 18.64
39

• high school seniors smoking

Mean 18.64
Avrg Moving Range 0.778
Upper Process Limit 20.71
Lower Process Limit 16.57
40
Part II. Patient Safety
41
Heinrich Triangle
Knowledge
42
Parallel Universe
43
Essential System Characteristics

• Uses available technologies
• Real-time data
• Feedback providing (closing the loop)
• Designed to succeed (safe)

44
ALCOA

• At ALCOA I have a real fine data system so that
  I knew every minute of every day the health and
  safety condition of 140,000 people.  We shared
  the information across the whole place so that we
  had real-time learning among the people.  The
  information was not there for me.  It was for
  140,000 people to learn from shared experiences.
  Without information having to travel up through
  some appointment process and maybe some day gets
  distributed so you can learn something.  It was
  there every day. 
• If we had an incident in Sumatra, the people in
  Jamaica knew it tomorrow morning and they did
  something about it to avoid the same kind of
  circumstances. 
• When I asked for the data at Treasury, it took
  them a long time to get it for me and when they
  did, it turned out that their lost workday rate
  in the Treasury, that has about the same number
  of employees, was 20 times higher than ALCOAs.
• Paul ONeill, Treasury Secretary

45
J.T. Reason

• major residual safety problems do not belong
  exclusively to either the technical or the human
  domains. Rather, they emerge from as yet little
  understood interactions between the technical and
  social aspects of the system
• J.T. Reasons,
• Safety at Sea and in the Air- Taking
  Stock Together Symp., Nautical Institute,
  1991

46
Disney

• But, ultimately, even the most conscientious
  Cast Members cannot do it alone. Guests, too,
  have an essential role to play in making every
  visit to our parks safe.
• Paul S. Pressler,
• Chairman, Walt Disney Parks and Resorts

47
Aviation Safety Network

• Without a doubt 2001 was the year with the
  highest aviation caused fatalities ever.
  However, when we take a closer look at the
  figures we can see that 34 fatal multi-engined
  airliner accidents were recorded, which was an
  all-time low since 1946.

48
Learning Objectives

• Implement a blame-free reporting culture
• Improve or expand chemotherapy taxonomy/definition
  s
• Preventable adverse drug events and near misses
• Operational barriers (i.e., delays) as errors?
• Evaluate wireless technologies as an electronic
  resources and reporting tool
• Integrate into daily workflow
• Extend to bedside
• Apply Computerized Order Entry/Decision support
• Quality Improvement via multidisciplinary
  teamwork based on data feedback

49
Intelligent Chemo Delivery System
50
Chemo Events Data Capture
51
Standardized Reporting
52
Lesson Learned

• Leadership and organizational culture are as
  critical for patient safety as structure
• Vertical and horizontal organization
  communication are essential components of
  surveillance, prioritization
• Team communication (chatter) is key to developing
  safe culture and trust Foundation for safety
  pattern recognition
• Timely data feedback drives safety improvement
• Healthcare can learn much about systems thinking
  from other industries and cultures
• Tightly coupled systems are more prone to failure
  than highly adaptive systems

53
Student Project

• Develop and test taxonomy for systems errors in
  Emergency Medicine
• What system factors in the ER increase the
  likelihood that care providers will commit
  errors?
• How to measure these factors?
• Clinic Redesign - Orthopedics
• Room Utilization tracking program