Tees Evidence Based Practice Course

1
Tees Evidence Based Practice Course

Understanding variation

• The James Cook University Hospital Thursday 28th
  July 2005
• Tony Roberts
• Head of Clinical Effectiveness Honorary
  Research Fellow
• North Tees Primary Care Trust, South Tees
  Hospitals Trust
• The University of Durham
• Tees Evidence Based Practice Group
• Amanda McGough, Anne Lister
• www.teesebp.net

2
Statistical Process ControlUnderstanding
variation...

• Write a line of 10 as using the wrong hand,
  ie not the one you normally write with
• Take as your model
• a

3
Ranking performance

• Look at your row of as
• Are some better than others? Place them in a rank
  order (top and bottom of the list is easy)
• What did you do wrong when you did the bad ones?
• What did you do right when you did the good ones?

4
Ranking can be unhelpful

• The row was created by the same person, in the
  same place, at the same time, with the same pen,
  the same paper and using a model to copy from
• Yet variation is ever present, the as are not
  the same (except to common sense)
• But it is absurd to ask what you did differently.
  The variation is due to the complexity intrinsic
  to the process of writing

5
Variation in a system is normal

• The variation is caused by factors that are
  inherent in the system over time
• They affect all outcomes
• This is common cause variation or
• The causes are unassignable

6
More variation...

• Now draw a row of as with your correct hand
• Take as your model
• a

7
Common cause variation can be reduced

• For most people it is obvious that this row of
  as is different (and better) to the row written
  with the wrong hand
• So common cause variation can be reduced by
  tackling things that affect the process all the
  time

8
A different sort of variation...

• Now draw a row of as with your correct hand,
  but somewhere along the row, do one a with the
  wrong hand
• Show it to a neighbour (who didnt see you do it)
  and see if they can tell which is the wrong
  a.
• When variation affects a pattern enough, there
  may be more than common causes operating

9
Some variation may not be normal

• The factors are not present in the process all
  the time
• They do not affect everybody
• They arise because of specific circumstances
• This is special or assignable cause variation.

10
What to do if there is anassignable cause

• 95 of failures are due to the system
• Dont wait until the trail is cold
• Check the data (again)
• Check case-mix
• Check facilities staffing
• Check the processes
• Check the individual

11
How does this help?

• Performance data should be used to guide quality
  improvement
• The purpose should be to find the assignable
  causes and understand their origin they should
  be prevented if bad and spread if good
• When only unassignable causes are present, the
  process can only be improved by changing things
  that affect the process all of the time

12
Does this help?

• League tables with only common cause variation
  tend to encourage unwarranted tampering
• League tables may lead to local special cause
  variation being ignored
• League tables tend to encourage the blame
  culture and are not linked directly to
  improvement activity

13
How much variation is normal?

• There is a trade-off between failing to spot real
  differences and falsely claiming them
• Limits should be locally agreed and are commonly
  set at 3 sigma
• For the statistically minded, sigma is the same
  as standard deviation, but calculated using
  formulas that take account of the order of the
  data

14
Two types of SPC chart

• If you want to compare a single individual, unit
  or hospital over different time periods, atime
  chart may be helpful
• If you want to compare different individuals,
  units or hospitals etc over a single time period,
  a funnel chart may be helpful

15
Anatomy of an SPC time chart
16
Anatomy of an SPC funnel chart
Special cause
17
Conclusion

• If you want to learn more, courses are beginning
  to appear (eg Birmingham University)
• The Modernisation Agency are using SPC and new
  CHAI are also looking at it
• The best book is Raymond Careys Improving
  Healthcare with Control Charts (2003) but is not
  for the fainthearted
• Download copies of the Excel spreadsheets for
  producing charts SPC Builder will make funnel
  plots and is excellent. Time-based charts are
  good in Excel, and we have a spreadsheet via
  Davis Balestracci.
• If you want software to do it for you, called
  WinChart, you can get a free trial copy by
  emailing winchart_at_ntlworld.com