Monitoring Adverse Events in Surgery

1
Monitoring Adverse Events in Surgery

• Steve Gallivan
• Clinical OR Unit
• University College London

2
Principal Collaborators

• Jocelyn Lovegrove (Ex-CORU)
• Chris Sherlaw-Johnson (CORU)
• Tom Treasure (Cardiac Surgeon)
• Jaroslav Stark (Paediatric Cardiac Surgeon)
• Marc de Leval (Paediatric Cardiac Surgeon)

3
Typical CUSUM plot
4
Factors contributing to model of surgical risk
5
Cumulative perioperative mortalities
10
Expected mortality (from risk model)
8
Actual mortality
Par for the
6
course
4
Net life
2
gain
0
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
Operation number
Calculating Net Life Gain based on pre-operative
risks
6
VLAD plot for a single surgeon
7
Unexpected
Net life gain
5
death
Surgeon A
4
Surgeon B
3
Surgeon C
2
1
Survivor
0
against
-1
the odds
-2
-3
-4
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
Operation number
Comparing three fictitious surgeons
8
(No Transcript)
9
6 more survivors than
predicted for the case mix.
Within 90 interval
Displaying probability that divergence a result
of chance
10
Net lives saved
VLAD plot for a professor of cardiac surgery
11
Comparing several surgeons at a single hospital
12
VARIABLE LIFE ADJUSTED DISPLAY - CABG
VLAD plot comparing Bristol with another cardiac
centre
13
Difficulties using VLAD for paediatric cardiac
surgery

• Many different types of procedure
• Operation may involve several procedures
• Surgeons perform relatively few operations of the
  same type
• Mortality standards not established
• No accepted risk scoring system

14
Monitor performance Compare surgeons
Change emphasis
Early identification of periods of divergent
outcome
15
Mortality rates for 11 paediatric cardiac
surgeons (1 year data)
16
Comparator Centres Mortality Rates
17
Surgeons own mean mortality
Partial risk strata
Surrogate pre-operative risk forecast
Risk model tailored to surgeons own mean
mortality rate
18
Normalised VLAD for a single surgeon
19
Complexity Category
20
Complexity Profile
100
80
Tom
Cumulative Percentage
60
Dick
40
Harry
20
0
1
2
3
4
5
6
Operation Difficulty
21
(No Transcript)
22
Case study 2
A heart transplant centre audits its recent
outcome and discovers that there have been 5
deaths out of the most recent 14
operations. Should the service be suspended?
23
KEY QUESTION
What is probability that a binary sequence length
M has a sub-string length N with at least K
ones? Probability that i-th bit is one qi
SURELY FELLER ANSWERED THIS!
24
N Most recent 6-string Number of ones

1 0
0 2 01
1 3 010
1 4 01001
2 5 010010
2 6
0100101 3 7
01001011 3 8
010010110 3
9 0100101101
4 10 01001011010
3 11 010010110101
4 12 0100101101011
4 13 01001011 010111
4 14 010010110101111 5
EVOLUTION OF BINARY STRINGS
25
Probability of a subsequence length M with K
deaths
26
EXAMPLE FOR 3 DEATHS OUT OF A SUBSEQUENCE OF 4
27
Recurrence relationship for evolution of
probabilities for binary string b from the set HMK
f1(b) and f2(b) the two progenitors of b
28
Probability of a run with 5 deaths out of a
subsequence of 14
29
Probability of poor run when mortality is 16