American Recovery and Reinvestment Act

Epidemiology and Laboratory Capacity (ELC)

for Infectious Disease Program
Healthcare-Associated Infections (HAIs)
G
Grantee
M
Meeting
i

Jonathan R. Edwards, M.Stat

October 19-20, 2009

NHSN HAI Comparison

p
Metrics
Jonathan R.
R Edwards
Edwards, MStat
Statistics Team Leader, DHQP

Obj ti
Objectives

Review HAI point estimation

Explain standardized infection ratio (SIR)

Describe improved SIR calculations

Discuss options for hospital comparisons

Point Estimation for

HAI Surveillance

A type of statistical inference often

resulting in the calculation of HAI rates
Turn surveillance data into decision
making information
information
Allows you to express surveillance
information in numerical terms or metrics
Numerator
Denominator

X Constant

I id
Incidence
Densit
ity Ratte

Type of rate that accounts for differences in follow-up

(exposure) time for each person included in the at-risk

at risk
population.
Numerator is number of new cases of disease.
disease.

Denominator is person-time units.

Example 1:
Central line-associated BSI (CLABSI) rate
12 CLABSIs
000 ce
central
t a line-days
e days
4000

X 1000 = 3.0

Example
a peC
CLABSI
S Rate
ate Table
ab e

Aggregate CLABSI Rate

Local CLABSI Rates

p-values for local vs. aggregate

Percentile rank among all units

P
Proportion
ti
A fraction in which the numerator is included
within the denominator
Numerator is the magnitude of the part
Denominator is the magnitude of the whole
Often expressed as a percent

Example 2:
Surgical site infection rate:
4 SSI in hysterectomy patients
280 women undergoing hysterectomy

X 100 = 1.4%

Example SSI Rate Table

Local SSI Rates

Aggregate SSI Rate

p-value
l for
f local
l
l vs. aggregate
t

Percentile rank among all hospitals

Standardized
Infection Ratio
(SIR)

St d di d IInfection
Standardized
f ti R
Ratio?
ti ?

Standardized Infection Ratio, SIR, is a

summary measure used to compare the
HAI experience among one or more groups
off patients
ti t tto th
thatt off a standard
t d d populations
l ti
(e.g. NHSN)

I di t standardization
Indirect
t d di ti method
th d

Accounts for differences in risk of HAI

among the groups

C l l ti an SIR
Calculating
Observed (O) HAIs
SIR =
Expected (E) HAIs

To calculate O, sum the number of HAIs among a

group
T calculate
To
l l t E,
E requires
i
the
th use off the
th appropriate
i t
aggregate data (risk-adjusted rates)

Central Line
Line-associated
Bl d t
Bloodstream
Infections

A
Annual
l NHSN R
Reportt

Edwards
d a ds J
JR et a
al. Am J Infect
ect Co
Control
t o 2008;36:609-26.
008;36 609 6

A
Annual
l NHSN R
Reportt

Edwards
d a ds J
JR et a
al. Am J Infect
ect Co
Control
t o 2008;36:609-26.
008;36 609 6

E
Example
l CLABSI R
Rate
t T
Table
bl
Type of
Location

#
# Central CLABS NHSN p-value
CLABSI line-days
line days
I
Rate
Rate

Coronary

380

5.26

2.1

0.09

Cardiothoracic

257

3.89

1.4

0.15

Medical

627

4.78

2.4

0.11

Med/Surg,
major teaching

712

2.81

2.0

0.32

T t l
Total

1976 4.05
4 05

---

---

E
Example
l CLABSI R
Rate
t T
Table
bl
Type of
Location

#
# Central CLABS NHSN p-value Expected
CLABSI line-days
I
Rate
# of
CLABSI
R t
Rate

Coronary

380

5.26

2.1

0.09

Cardiothoracic

257

3 89
3.89

14
1.4

0 15
0.15

Medical

627

4.78

2.4

0.11

Med/Surg
Med/Surg,
major teaching

712

2 81
2.81

20
2.0

0 32
0.32

Total

1976 4.05

---

0.8

---

Expected Number = 380 * (2.1 / 1000) = 380 * 0.0021 = 0.8

E
Example
l CLABSI R
Rate
t T
Table
bl
Type of
Location

#
# Central CLABS NHSN p-value Expected
CLABSI line-days
I
Rate
# of
CLABSI
R t
Rate

Coronary

380

5.26

2.1

0.09

0.80

Cardiothoracic

257

3 89
3.89

14
1.4

0 15
0.15

0 36
0.36

Medical

627

4.78

2.4

0.11

1.15

Med/Surg
Med/Surg,
major teaching

712

2 81
2.81

20
2.0

0 32
0.32

1 42
1.42

Total

1976 4.05

---

---

3.73

E
Example
l CLABSI R
Rate
t T
Table
bl
Type of
Location

#
# Central CLABS NHSN p-value Expected
CLABSI line-days
I
Rate
# of
CLABSI
R t
Rate

Coronary

380

5.26

2.1

0.09

0.80

Cardiothoracic

257

3 89
3.89

14
1.4

0 15
0.15

0 36
0.36

Medical

627

4.78

2.4

0.11

1.15

Med/Surg
Med/Surg,
major teaching

712

2 81
2.81

20
2.0

0 32
0.32

1 42
1.42

Total

1976 4.05

Standardized Infection Ratio ((SIR)) =

---

---

3.73

Observed
= 2.14

E pected= 3.73
Expected
3 73
8

External Comparison of an SIR to

its Nominal Value
Val e of 1
1.0
0
SIR = 1.0

Observed Number of SSI

= Expected Number of SSI

Is the SIR of Hosp A (2.14) significantly greater than 1.0?

Answer: p-value = 0.07

(Poisson test)

Conclusion: NO,
NO the observed number of CLABSI is not
"significantly greater" than expected

Surgical
g
Site
Infections

Annual NHSN Report

Edwards JR et al. Am J Infect Control 2008;36:609-26.

Annual NHSN Rep

port

Edwards JR et al. Am J Infect Control 2008;36:609-26.

E
Example
l SSI Rate
R t T
Table
bl
Procedure

Index

Cardiac surgery

CBGB

#
SSI

#
procedures

SSI
Rate

0,1

80

3.75

2.02

0.25

2,3

20

15.0

5.29

0.08

10

10.0

1.59

0.16

10

230

4.35

3.15

0.20

60

8.33

5.76

0.28

2,3
T t l
Total

22

400 5.50
5 50

NHSN
Rate

---

p-value

---

E
Example
l SSI Rate
R t T
Table
bl
Procedure

Index

Cardiac surgery

CBGB

#
SSI

#
procedures

SSI
Rate

0,1

80

3.75

2.02

0.25

2,3

20

15.0

5.29

0.08

10

10.0

1.59

0.16

10

230

4.35

3.15

0.20

60

8.33

5.76

0.28

2,3
T t l
Total

22

400 5.50
5 50

NHSN
Rate

---

p-value

Expected
# of SSI

---

Expected Number = 80 * (2.02 / 100) = 80 * 0.0202 = 1.6

1.6

E
Example
l SSI Rate
R t T
Table
bl
Procedure

Index

Cardiac surgery

CBGB

#
SSI

#
procedures

SSI
Rate

0,1

80

3.75

2.02

0.25

1.6

2,3

20

15.0

5.29

0.08

1.1

10

10.0

1.59

0.16

0.2

10

230

4.35

3.15

0.20

7.2

60

8.33

5.76

0.28

3.5

2,3
Total

22

400 5.50
5 50

NHSN
Rate

---

p-value

---

Expected
# of SSI

13 6
13.6

E
Example
l SSI Rate
R t T
Table
bl
Procedure

Index

Cardiac surgery

CBGB

#
SSI

#
procedures

SSI
Rate

01
0,1

80

3 75
3.75

2 02
2.02

0 25
0.25

16
1.6

2,3

20

15.0

5.29

0.08

1.1

10

10.0

1.59

0.16

0.2

10

230

4.35

3.15

0.20

7.2

60

8.33

5.76

0.28

3.5

2,3
Total

22

400 5.50

NHSN
Rate

---

Observed
Standardized Infection Ratio (SIR) =
Expected
=

p-value

---

22

Expected
# of SSI

13.6

13.6
13 6

= 1.62

Standardized Infection Ratio (SIR)

Ratio of Observed to Expected

infections

Risk-adjusted summary measure

Used to compare overall HAI rates of

any two patient cohorts, groups, or
hospitals

External Comparison of
an SIR to its Nominal Value of 1.0
SIR = 1.0

Observed Number of SSI

= Expected Number of SSI

Is the SIR of Team A (1.62) significantly greater than

1.0?
Answer: p
p-value
value = 0
0.01
01

(Poisson test)

Conclusion: YES,
YES the observed number of SSI is
"significantly greater" than expected

C fid
Confidence
IInterval
t
l ffor SIR

A (1 )100% confidence interval for p

Z
O 1 1 2

SIRL = E 9O
9O

1
(O + 1) 1
Z 2

+
SIRU =

E 9(O + 1)
9(O + 1)

Example normal curve values

90% confidence z.05 = 1.645
95% confidence z.025 = 1.96
99% confidence z.005 = 2.58

Liddell FDK. JECH 1984;38:85-89.

Advances in Risk-Adjustment:
Risk Adjustment:
Comparing Surgical Site Infections
using a Standardized Infection Ratio

Skip

Logistic Regression Analysis

Incisional SSI after CSEC
Factor
Parameter Estimate
Intercept
-6.300
Body Mass
0.046
Duration of labor 9 hr
0 506
0.506
ASA 3
0.598
p
age
g (years)
(y
)
14.961
Reciprocal

OR
1.047
1 659
1.659
1.818
999.0

p-value
<0.0001
<0 0001
<0.0001
<0.0001
0.0018

* Horan T. et al. Risk Factors ffor Incisional S

Surgical S
Site Infection
f
after
f
Cesarean Section: Results of a 5-year Multicenter Study, ICHE 2000;
S-T4-05

L i ti R
Logistic
Regression
i M
Model
d l

General formula
logit (p) = + 1 X1+ 2 X2 + 3 X3 + .
Where:
p probability
b bilit off SSI after
ft operation
ti
intercept
1, 2 , 3 ... parameter estimate
X1, X2 , X3 ... risk factors

Logistic Model
for CSEC:

Factor
Parameter Estimate
Intercept
-6.300
B d M
Body
Mass
0 046
0.046
Duration of labor > 9 hr
0.506
ASA > 3
0.598
Reciprocal age (years) 14.961

logit (p) = + 1 X1+ 2 X2 + 3 X3 + 4 X4

= -6.300 + 0.046 (Body Mass) +
0.506 (Duration of labor 9 hr*)
hr )
+

0.598 ((ASA 3*)) +

14.961 ( 1 / Age)
*For
For these risk factors
factors, if present = 1; if not = 0

Example Patient #1
Risk Factors

Body mass = 28
Duration of labor = 12 hours
ASA score = 1
Age = 22

Logistic
g
Model
Calculation for Example Patient #1
logit(
g ( p ) = + 1 X1+ 2 X2 + 3 X3 + 4 X4
logit( p ) = -6.300 + 0.046 (28) + 0.506 (1) +
0 598 (0) + 14.961
0.598
14 961 (1/22) = -3.825
3 825

Solve for p : p = e logit (p ) / (1 + e logit (p) )

p = e -3.825 / (1 + e -3.825)= 0.022 or 2.2% risk

Example Patient #2
Risk Factors

Body mass = 36
Duration of labor = 9 hours
ASA score = 3
Age = 18

Logistic
g
Model
Calculation for Example Patient #2
logit
g ( p) = + 1 X1+ 2 X2 + 3 X3 + 4 X4
logit (p
) = -6.300 + 0.046 (36) + 0.506 (1) +
0 598 (1) + 14.961
0.598
14 961 (1/18) = -2.708
2 708

:
Solve for p

= e logit (p ) / (1 + e logit (p ))

p = e -2.708 / (1 + e -2.708)= 0.063

or 6.3% risk

List of Patient Risk Factors

Incisional SSI after CSEC
Body
Patient Mass Duration ASA Age SSI
of labor
1
28 12
1
22
0
0.022
2
36 9
3
18
1
0.063
3
30 7
2
28
0
0 012
0.012
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
100
8
4
38
0
0.020
31
Total
O = 3 E = 2.3
23
SIR = O/E = 3 / 2.3 = 1.3

I t
Interpreting
ti an SIR

If SIR = 1, or O = E, observed SSIs equals

the expected SSIs based on NHSN

If SIR is significantly >1, this suggests the

institution may need further investigation
for the contributing procedures

C
Conclusions
l i

Appropriate multivariate models are useful to

b tt estimate
better
ti t SSI risk
i k

Calculation of a risk-adjusted Standardized

Infection Ratio (SIR) will improve
comparisons
i
ffor h
hospitals
it l

H
Hospital
it l SIR
SIRs C
Compared
d tto 1
Hospital

SIR

1.2

p-value Status
group
0.12
Same

09
0.9

0 23
0.23

Same

2.7

0.001

High

0.7

0.002

Low

15
1.5

0 001
0.001

High

SIR (95
5% CI)

H
Hospital
it l SIR G
Groups
2

reference
f
liline

F G

Hospitals

SIR Funnel Plots

Standa
ardized In
nfection rratio

E
Exact
t SIR lilimits
it

Volume
ou eo
of cases

P t/P
Past/Present
t Hospital
H
it l SIR
SIRs
Hospital
A

Past
SIR
1.2

Present Status
SIR
group
1.6

Same

0.9

0.8

Same

2.7

1.9

Lower

0.7

1.1

Higher

15
1.5

12
1.2

Lower

Thank You