improvement science in action -...
TRANSCRIPT
4/17/2014
1
Improvement Science In Action
A Run Through the Run Chart:
Using Run Charts for Improvement
Sandra Murray
May 1st, 2014
Session Objectives
By the end of this session you will
be able to:
Identify what a run chart is
Know when to start a run chart
Be able to interpret a run chart
Identify several keys to effective ways to display data
with a run chart
Be aware of cautions and special uses for run charts.
DG p. 67
4/17/2014
2
What are we trying to accomplish?
How will we know that a change is an improvement?
What change can we make that will result in improvement?
Model for Improvement
Act Plan
Study Do
Used with permission: Improvement Guide Improvement Guide, Jossey-Bass, 2009
Going
Deeper
Run Chart
Graphical display of data plotted in some type of
order. Also has been called a time series or a
trend chart.
DG p. 67
4/17/2014
3
We Can Learn a Lot with a Simple
Tool…
What do you observe here? What would your question
be?
DG P. 68
What are Classic Uses of the Run
Chart?
Displaying data to make process performance visible.
Determining whether a change resulted in improvement
Determining whether gains made are sustained
DG p. 68
4/17/2014
4
DG p. 70
4/17/2014
5
4/17/2014
6
Determining the Median
4/17/2014
7
Analyzing a Run Chart Starts with a simple visual analysis
They are testing changes here. Do they have
improvement yet? Charts speak for themselves.
Analysis is by degree of belief
4/17/2014
8
Unplanned Returns to OR: Pilot Population
0
1
2
3
4
5
6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
%
Months
Pre-Procedural Briefings
Razors to Clippers
Bleeding risk assessment, DVT Proph
Beta Blocker use,
Normothermia
(N~200/Mo.) Good
Prophylactic ABX Timing
Improvement Projects Require a Family of Measures
• 2-8 measures typically -Each on a graph -All viewed on one page
DG p. 74
Fig 3.6: Improvement Evident Using a Set of Run Charts Viewed on One Page
4/17/2014
9
Small Multiples
Multiple run charts viewed on one page
All these run charts are about the same measure but for
a different location, provider or segment of the
population
Each has the same scale vertically and horizontally
Allows for rapid comparison
DG p. 75
4/17/2014
10
Average
Waiting Times:
All Primary Care
Clinics in VHA
System
Small Multiples:
Overall System
and 22 Districts
May Display More Than One
Measure on a Graph
DG p. 75
4/17/2014
11
May Use Different Measure for
Each Axis
DG p. 76
Sometimes We Don’t Have Much Data
May not be rich in data but that data may still lead to a high degree
of belief in the change(s) tested
Characterize the change by describing the before and after medians
Minimizes point-to-point variation
DG p. 77
4/17/2014
12
Probability Based Rules for Run Chart Analysis
If visual analysis leaves us uncertain that change(s)
yielded improvement we may use probability based rules
to analyze the run chart
DG p. 77
DG p. 78
4/17/2014
13
Rule 1: Shift
Six or more consecutive POINTS either all above or all below the median. Skip values on the median and continue counting points. Values on the median DO NOT make or break a shift.
Median=10 Median=11
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25Me
as
ure
or
Ch
ara
cte
ris
tic
Rule 1
Median 10
Rule 2: Trend
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Me
as
ure
or
Ch
ara
cte
ris
tic
Rule 2
Median 11
• Five points all going up or all going down. If the value
of two or more successive points is the same count the
first one then ignore the identical points when counting;
like values do not make or break a trend.
4/17/2014
14
Rule 3:Runs
– A run is a series of points in a row on one side of the median. Some points fall right on the median, which makes it hard to decide which run these points belong to.
– So, an easy way to determine the number of runs is to count the number of times the data line crosses the median and add one.
– Statistically significant change signaled by too few or too many runs.
DG p. 81
Rule 3: # of Runs Table for Checking for Too Many or Too Few Runs on a Run Chart
Total number of data
points on the run chart
that do not fall on the
median
Lower limit for the number of runs
(< than this number of runs is “too few”)
Upper limit for the number of runs
(> than this number of runs is “too many”)
10 3 9
11 3 10
12 3 11
13 4 11
14 4 12
15 5 12
16 5 13
17 5 13
18 6 14
19 6 15
20 6 16
21 7 16
22 7 17
23 7 17
24 8 18
25 8 18
Table is based on about a 5% risk of failing the run test for random patterns of data. Frieda S. Swed and Churchill Eisenhart,
(1943). “Tables for Testing Randomness of Grouping in a Sequence of Alternatives. Annals of Mathematical
Statistics. Vol. XIV, pp.66 and 87, Tables II and III
DG p. 80
4/17/2014
15
Rule 3: Runs
– Too many runs
– Not usually indicative of improvement
– Most likely data that needs to be stratified or another data collection or display problem
DG p. 81
RULE 4
For detecting unusually large or small numbers:
Data that is Blatantly Obvious as a different value
Everyone studying the chart agrees that it is unusual
Remember: – Every data set will have a high and a low - this does not mean the
high or low are astronomical
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Mea
sure
men
t or
Cha
ract
eris
tic
Rule 4
4/17/2014
16
What is Our Job?
When we see signal:
– Go learn and take appropriate action
– Is the signal an improvement?
– If so how keep it, spread it?
– Is signal bad news?
– If so, how get rid of it?
– How do we prevent it?
Exercise 1: Analyze This Chart
%
Percent Incomplete Revised Trauma Score in Record14.8 13.0 13.6 14.0 13.4 13.3 13.8 13.7 14.0 14.4 13.1 14.1 14.8 13.4 13.7 14.1 12.9 9.7 10.9 11.6 10.5 10.3 9.8 10.2 9.9%
Median = 13.4
Test 1
Test 2
Test 3
Test 4
Implementation
Jun 07 J A S O N DJan 08 F M A M J J A S O N DJan 09 F M A M J
8
10
12
14
16
18
Percent Unplanned Readmissions
4/17/2014
17
Exercise 2: Analyze This Chart
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10 11 12 13
% Reenrolling W/O a Gap
Median = 19
Weeks
Ad campaign
started
Exercise 3: Analyze This Chart
0
5
10
15
20
25
30
35
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
# Referrals to BH Services
#
Weeks
Median =
4/17/2014
18
Exercise 4: Analyze This Chart
Fundamental Uses of Run Charts
Display data to make process performance visible
Determine whether a change resulted in evidence of
improvement
Determine whether we are holding the gain made by our
improvement
4/17/2014
19
DG p. 86
When Do We Start a Run Chart?
DG p. 87
4/17/2014
20
Proper Use of the Median
When should we apply a median?
– Will depend on your situation
– If very little data baseline median may be only a few data points
– If want to apply probability-based rules for analysis of run chart
need 10 data points for median
– If graph shows no signals (shift, trend, runs astronomical) and
median made from 10 or more data points freeze and extend
median into the future
– This will result in earliest possible detection of signals
DG p. 87
•If median not frozen and extended will result in
delayed detection of signals
DG p. 89
4/17/2014
21
•A signal is detected utilizing both original and extended
median
DG p. 89
•If a signal is detected and sustained a new median may be
created for the new process performance
-When analyzing run chart with two separate medians rules are
must be applied separately to the data surrounding each median
DG p. 90
4/17/2014
22
Plotting Rare Events
Problems:
– Results in too many zeros
– Makes interpretation difficult and chart of little
value
– Useful alternative is to chart time or workload
between undesirable events
– Up is always good for these charts
DG p. 91
DG p. 92
4/17/2014
23
DG p. 92
DG p. 94
4/17/2014
24
Cautions with Percent or Rate Data
on Run Chart Run chart not sensitive to size of denominator
– If denominator unusually small or large compared to other
denominators used for that chart may result in misleading chart
– For an effective run cart denominators for each data point should
be +/- 25% of average denominator size
Specific types of Shewhart control charts will cope with
unequal denominator sizes
DG p. 93
DG p. 94
4/17/2014
25
Data Line Guidance
May connect data points with a line if data in
time order
If data not in time order do not connect data
points
DG p. 95
Trend Lines on Run Charts
Place only if detect signal on run chart
DG p. 97
4/17/2014
26
No signal of improvement on chart
Improper to use trend line
DG p. 98
Stratification
The separation and classification of data according to
selected variables or factors
– Unit, patient type, clinic, service, age group, diagnosis, etc.
My give us clues as to where to focus or ideas for
change to test
DG p. 99
4/17/2014
27
DG p. 100