Using Primary Care Data for Quality Improvement

1
Using Primary Care Data for Quality Improvement

• Dr John Derry
• Primary Care Medical Adviser Thames Valley SHA

2
Preview

• Examples of clinical audit data
• Similar to QMAS data
• Understanding variation
• The role of SPC
• Using SPC methods to interpret clinical audit data

3
CHD Audit Resultsage sex standardised
prevalence of CVD
4
CHD Audit Resultsblood pressure recording
control
5
So what now?

• See variation
• What is significant?
• Is it OK to be near average?
• When should we act?
• How should we act?
• Which data can we use?

6
Use of SPC

• Statistical Process Control
• Methods developed by Shewhart and Deming (1930s
  - 1990s)
• Cornerstone of quality improvement
• Two different kinds of variation can affect any
  process
• Distinguish by statistical methods

7
Statistical Process Control

• First and foremost, a way of thinking with some
  tools attached
• About the continual improvement of processes and
  outcomes
• About getting the most from your processes
• Quotes from Don Wheeler in Understanding
  Variation SPC Press, 2000

8
Two types of variation

• How long does it normally take you to get to
  work?
• Why does it vary?
• How do you use this understanding to plan your
  journey?
• When to leave the house
• Which route to take
• When to make a change

9
Understanding variation

• Routine
• common causes
• many factors, some unknowable
• noise in the system
• affects process most of the time
• part of the process
• variation is predictable

• Exceptional
• special causes
• assignable causes
• usually few, not many
• can usually be identified
• not part of the process
• intermittently apparent
• unpredictable variation

10
What to do about variation

• Exceptional
• investigate each point outside the limits
• look for the special cause and do something about
  it
• almost always something to find
• opportunities to learn

• Routine
• dont react to individual results
• look at the average and process limits
• improve the whole process if these not acceptable
• or continuously improve quality!

11
Two kinds of mistake

• Mistake 1
• Act as if there is a special cause when there is
  only routine variation
• Might make things worse
• Wasted effort anyway

• Mistake 2
• Fail to spot a special cause assume there is
  just routine variation present
• Missed opportunity
• reduce variation
• improve quality
• learn something

12
Control charts

• Graphical method developed by Shewhart to help
  distinguish these two kinds of variation
• routine and exceptional
• predictable and unpredictable
• common and special cause
• Process behaviour charts(Don Wheeler)

13
An example control chart
routinevariation
14
Another type of control chart
Control Chart of Clinical Audit Data
25
20
15
SqRt number with criterion
Routine variation
10
5
0
0.00
20.00
40.00
60.00
80.00
100.00
SqRt number without criterion
15
How to interpret the chart
Control Chart of Clinical Audit Data
25
20
15
Practices here cannot be distinguished from
average
SqRt number with criterion
10
5
0
0.00
20.00
40.00
60.00
80.00
100.00
SqRt number without criterion
16
Double square-root chart

• Described recently by Mohammed et al. (Lancet
  2001 357 463-67)
• Originally developed by Fisher, Tukey Mosteller
  in 1940s
• Enable analysis of variation in cross-sectional
  data
• Based on binomial probability distribution for
  calculating SD

17
Types of control chart

• For measurement (variable) data
• Single observations
• Limits based on average moving range
• Average /- (3/bias correction factor d2) x
  average MR
• Subgroups of observations
• Limits based on std deviation of subgroup
• Correction factor depends on number in each
  subgroup

18
Types of control chart

• Count (attribute) data
• Yes/no, with/without data
• P-chart
• Limits based on Binomial conditions
• Average p /- 3 x sqrt (p(1-p)/n)
• event data count (needlestick injuries)
• U or C chart (denominator varies or constant)
• Limits based on Poisson conditions

19
Some real examples

• Using clinical audit data

20
Standardised CVD Prevalence
Average 4.6 /- 3SD Range 3.6-5.7
25.00
20.00
434
413
416
407
15.00
409
420
SqRt number with CVD
424
10.00
406
405
417
5.00
429
0.00
0.00
20.00
40.00
60.00
80.00
100.00
SqRt number without CVD
21
CHD Audit Resultsage sex standardised
prevalence of CVD
22
BP recorded
23
CHD Audit Resultsblood pressure recording
control
24
Audit ResultsCardiovascular Disease Prevalence
25
Audit ResultsCVD Patients with cholesterol record
26
Audit ResultsCholesterol levels in CVD Patients
27
Audit ResultsStatin Rx for CVD Patients
28
Control charts for clinical audit

• To answer the question
• What do we do now weve got the results?
• To identify where to target efforts
• To know when to act
• To know what kind of action to take

29
Issues to consider

• Using the right kind of chart for the data
• Time-series data is generally better
• Limitations of binomial charts
• Binomial conditions (are they met?)
• Probability of single item possessing the
  attribute is constant
• Each item is independent of others

30
Recommended Reading

• Improving Healthcare with Control Charts Basic
  and Advanced SPC Methods and Case Studies
• by Raymond G. Carey
• ISBN 0-87389-562-2
• American Society for Quality
• Quality Press, 2003
• www.asq.org Available at Amazon!