Perioperative cardiovascular mortality in noncardiac surgery: validation of the Lee cardiac risk index.

Abstract

The Lee risk index was developed to predict major cardiac complications in noncardiac surgery. We retrospectively evaluated its ability to predict cardiovascular death in the large cohort of patients who recently underwent noncardiac surgery in our institution.
The administrative database of the Erasmus MC, Rotterdam, The Netherlands, contains information on 108 593 noncardiac surgical procedures performed from 1991 to 2000. The Lee index assigns 1 point to each of the following characteristics: high-risk surgery, ischemic heart disease, heart failure, cerebrovascular disease, renal insufficiency, and diabetes mellitus. We retrospectively used available information in our database to adapt the Lee index calculated the adapted index for each procedure, and analyzed its relation to cardiovascular death.
A total of 1877 patients (1.7%) died perioperatively, including 543 (0.5%) classified as cardiovascular death. The cardiovascular death rates were 0.3% (255/75 352) for Lee Class 1, 0.7% (196/28 892) for Class 2, 1.7% (57/3380) for Class 3, and 3.6% (35/969) for Class 4. The corresponding odds ratios were 1 (reference), 2.0, 5.1, and 11.0, with no overlap for the 95% confidence interval of each class. The C statistic for the prediction of cardiovascular mortality using the Lee index was 0.63. If age and more detailed information regarding the type of surgery was retrospectively added, the C statistic in this exploratory analysis improved to 0.85.
The adapted Lee index was predictive of cardiovascular mortality in our administrative database, but its simple classification of surgical procedures as high-risk versus not high-risk seems suboptimal. Nevertheless, if the goal is to compare outcomes across hospitals or regions using administrative data, the use of the adapted Lee index, as augmented by age and more detailed classification of type of surgery, is a promising option worthy of prospective testing.

Full text

Page 1

CLINICAL RESEARCH STUDY
Perioperative cardiovascular mortality in noncardiac surgery:
Validation of the Lee cardiac risk index
Eric Boersma,aMiklos D. Kertai, MD,bOlaf Schouten, MD,cJeroen J. Bax, MD,f
Peter Noordzij, MD,bEwout W. Steyerberg,dArend F.L. Schinkel, MD,a
Marian van Santen,eMaarten L. Simoons, MD,aIan R. Thomson, MD,g
Jan Klein, MD,bHero van Urk, MD,cDon Poldermans, MDc
aDepartments of Cardiology,bAnesthesiology,cVascular Surgery,dClinical Decision Sciences, andeMedical Registration,
Erasmus Medical Center, Rotterdam;fDepartment of Cardiology, University Hospital Leiden, Leiden, The Netherlands;
andgDepartment of Anesthesiology, University of Manitoba, Winnipeg, Canada.
ABSTRACT
PURPOSE: The Lee risk index was developed to predict major cardiac complications in noncardiac
surgery. We retrospectively evaluated its ability to predict cardiovascular death in the large cohort of
patients who recently underwent noncardiac surgery in our institution.
METHODS: The administrative database of the Erasmus MC, Rotterdam, The Netherlands, contains infor-
mation on 108 593 noncardiac surgical procedures performed from 1991 to 2000. The Lee index assigns 1
point to each of the following characteristics: high-risk surgery, ischemic heart disease, heart failure,
cerebrovascular disease, renal insufficiency, and diabetes mellitus. We retrospectively used available infor-
mation in our database to adapt the Lee index calculated the adapted index for each procedure, and analyzed
its relation to cardiovascular death.
RESULTS: A total of 1877 patients (1.7%) died perioperatively, including 543 (0.5%) classified as cardio-
vascular death. The cardiovascular death rates were 0.3% (255/75 352) for Lee Class 1, 0.7% (196/28 892)
for Class 2, 1.7% (57/3380) for Class 3, and 3.6% (35/969) for Class 4. The corresponding odds ratios were
1(reference),2.0,5.1,and11.0,withnooverlapforthe95%confidenceintervalofeachclass.TheCstatistic
for the prediction of cardiovascular mortality using the Lee index was 0.63. If age and more detailed
information regarding the type of surgery was retrospectively added, the C statistic in this exploratory
analysis improved to 0.85.
CONCLUSION: The adapted Lee index was predictive of cardiovascular mortality in our administrative
database, but its simple classification of surgical procedures as high-risk versus not high-risk seems
suboptimal. Nevertheless, if the goal is to compare outcomes across hospitals or regions using administrative
data, the use of the adapted Lee index, as augmented by age and more detailed classification of type of
surgery, is a promising option worthy of prospective testing.
© 2005 Elsevier Inc. All rights reserved.
KEYWORDS:
Surgery;
Risk;
Prediction;
Database;
Cardiovascular
Requests for reprints should be addressed to Eric Boersma, Clinical
Epidemiology Unit, Department of Cardiology, Room H543, Erasmus MC,
Dr Molewaterplein 40, 3015 GD Rotterdam, The Netherlands.
E-mail address: h.boersma@erasmusmc.nl.
0002-9343/$ -see front matter © 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.amjmed.2005.01.064
The American Journal of Medicine (2005) 118, 1134-1141

Page 2

Patients undergoing major noncardiac surgery are at high
risk of cardiovascular morbidity and mortality. Evidence
suggests that perioperative myocardial infarction is com-
monly caused by rupture of a coronary plaque, leading to
thrombus formation and subsequent vessel occlusion, sim-
ilar to what occurs in nonsurgical settings.1,2The clinical
importance of perioperative cardiovascular complications is
well recognized, and numerous investigations have de-
scribed the relation between patient’s characteristics and the
risk of adverse cardiovascular outcome.3-11
In the 1990s, Lee and colleagues systematically analyzed
the risk of major cardiac complications (which included
myocardial infarction, cardiogenic pulmonary edema, car-
diac arrest, and cardiac death) in 4315 patients undergoing
noncardiac surgery, the largest cohort ever described.8The
resulting Lee index, which is a modification of the original
Goldman index, is now considered by many clinicians and
researchers to be the most relevant index to predict cardiac
risk. However, the patients studied by Lee and colleagues
cannot be considered as an average, unselected surgical
cohort because their study included only patients with a
several-day expected length of hospital stay and excluded
neurologic surgery; as a result, it was relatively dominated
by patients undergoing thoracic (12%), vascular (21%), and
orthopedic surgery (35%).
The current study evaluates the Lee index in 108 593
patients who underwent noncardiac surgery in the Erasmus
Medical Center from 1991 to 2000. Because we relied on
medical records and administrative data gathered as part of
routine medical care, rather than duplicating the prospective
methods of the Lee study, we used cardiovascular mortality,
instead of the broader range of clinical cardiac complica-
tions that were considered in the original Lee study, as our
primary endpoint.
Methods
Hospital setting, procedures and patients
The Erasmus Medical Center, a metropolitan university
hospital that serves a population of approximately 3 million
people in the southwestern area of The Netherlands, acts as
a tertiary referral center for approximately 30 affiliated
hospitals. Between January 1, 1991, and December 31,
2000, 122 860 noncardiac surgical procedures were per-
formed in patients above the age of 15 years in the Erasmus
Medical Center. We excluded 14 267 planned and un-
planned procedures that were conducted within 30 days
after an initial operation, and analyzed the perioperative
course of the remaining 108 593 procedures, which were
performed in 75 581 different patients. Over the 10-year
observation period, 20 885 patients had multiple surgeries
in the Erasmus Medical Center. They were included as
many times as they had surgeries that were more than 30
days apart. The median span between 2 successive proce-
dures was 297 days (interquartile range, 123 to 677 days).
The operation (not the patient) was the unit of analysis
because this approach is consistent with clinical practice,
wherein the risk of perioperative complications is assessed
in relation to a specific procedure.
Sources of data
For each patient undergoing surgery, a number of data items
are routinely entered into the computerized hospital infor-
mation system at the time the patient is hospitalized. First,
surgical techniques are classified by the treating physician
according to a standardized national coding system, which
was developed in cooperation with the National Health
Service and medical insurance companies. This system is
used for reimbursement and to record and monitor the
experience of surgeons and surgical residents. Using this
classification, we grouped surgical procedures into 14 cat-
egories, with a total of 11 969 procedures (11.0%) classified
into multiple categories.
Second, from written information that is provided by the
patient’s general practitioner, referring physician, or hospi-
tal physicians involved in perioperative care, each patient’s
medical history is classified according to the International
Classification of Diseases, Ninth Revision (ICD-9).12The
classification and the entry of the data in the electronic
database are performed by dedicated administrative person-
nel who have completed in-depth training on medical data
registration. We recorded the following medical conditions:
history of diabetes mellitus (ICD-9250), myocardial infarc-
tion (ICD-9410, 411, and 412), angina pectoris (ICD-9413
and 414), prior heart failure (ICD-9428), cerebrovascular
accident (ICD-9430), and renal disease (ICD-9580). These
recorded diagnoses served as surrogates for 5 more detailed
criteria used in the Lee index8in which ischemic heart
disease was defined as a history of myocardial infarction or
positive exercise test, current complaint of ischemic chest
pain, or use of nitrate therapy, or Q waves on the electro-
cardiogram, but patients with a history of coronary bypass
surgery or angioplasty were included only if they had cur-
rent complaints of chest pain presumed to be due to isch-
emia; heart failure was defined as a history of heart failure,
pulmonary edema, or paroxysmal nocturnal dyspnea, a
physical examination showing an S3gallop or bilateral
rales, or a chest radiograph showing pulmonary vascular
redistribution; renal failure was a creatine level ?2.0 mg/
dL; diabetes was insulin requiring; and cerebrovascular dis-
ease was defined as a history of a stroke or transient isch-
emic attack. For the score criterion, high-risk surgery, we
categorized procedures in the same way as the Lee index:
retroperitoneal, intrathoracic, or suprainguinal vascular pro-
cedures were defined as high risk.
Endpoint definition
The hospital information system also contains the patient’s
vital status at hospital discharge and clinical postoperative
1135Boersma et al Prediction of perioperative cardiovascular mortality

Page 3

diagnoses. For example, diagnosed perioperative myocar-
dial infarctions are reported, but the patients do not rou-
tinely have serial postoperative electrocardiograms or blood
sampling for determination of cardiac biomarkers. Conse-
quently, clinically unsuspected or painless perioperative
myocardial infarctions may well be missed. Similarly, clin-
ically apparent strokes were reported, but systematic neu-
rologic evaluations were not performed. In view of these
limitations, we chose cardiovascular death as the primary
endpoint of our analyses. Cardiovascular death, which was
defined as any death with a cardiovascular complication as
the primary or secondary cause (according to the definitions
of the World Health Organization), included deaths follow-
ing myocardial infarction, cardiac arrhythmia, resuscitation,
heart failure, or stroke. Noncardiovascular death was de-
fined as any death with a principal noncardiovascular cause,
including surgery-related bleeding complications, cancer,
trauma, and infection. Sudden death in a previously stable
patient was considered as cardiovascular.
To obtain the cause of death, 2 investigators (MDK, DP)
independently reviewed all available perioperative data but
were blind to preoperative characteristics and aimed to
reach consensus. If consensus could not be reached, the
opinion of a third, independent investigator (PN) was final.
Events were counted until hospital discharge or 30 days
after surgery, whichever day came first.
Data quality and ethical considerations
We should emphasize that the data that we used were
collected for administrative purposes and not for the pur-
poses of this study by clinicians using standardized data
collection forms. We designed and undertook this study
several years after the last patient was enrolled, and we were
not able to verify the completeness or the correctness of the
data. By necessity, we had to rely on the information that
was provided by the clinicians who took care of the patients
during everyday clinical practice.
This study was approved by the Medical Ethics Com-
mittee of the Erasmus Medical Center. However, given the
retrospective nature of our study, informed consent could
not be obtained from each patient.
Statistical analysis
Continuous data are described as median values and corre-
sponding 25th and 75th percentiles, and dichotomous data
are described as numbers and percentages. With the limita-
tions noted previously, we estimated the Lee index for each
patient in our dataset. Univariable logistic regression anal-
yses were used to evaluate the relation among the adapted
Lee index (with individual dummy variables for each score
category) as calculated with information in our administra-
tive database, the clinical characteristics that are part of this
index, and our primary endpoint, which was cardiovascular
death including stroke, rather than the endpoint of Lee and
colleagues, who considered the composite endpoint of myo-
cardial infarction, cardiogenic pulmonary edema, cardiac
arrest, or cardiovascular death. Crude, unadjusted odds ra-
tios and corresponding 95% confidence intervals are re-
ported.
Subsequently, multivariable logistic regression analyses
were performed to evaluate if the predictive power of the
adapted Lee index could be improved retrospectively in our
dataset by adding age or more detailed information on the
type of surgery, according to the classification recom-
mended by the American Heart Association/American Col-
lege of Cardiology.13The 4 categories were low risk (breast,
carotid, dental, endocrine, eye, gynecology, reconstructive),
low-intermediate risk (orthopedic, urologic), intermediate-
high risk (abdominal; ear, nose, throat; neurologic; pulmo-
nary; renal transplant; vascular, excluding aortic and ca-
rotid), and high risk (aortic). Adjusted odds ratios and
corresponding 95% confidence intervals were calculated.
The performance of risk models was determined by the C
statistic, which indicates how well a model rank orders
patients with respect to their outcomes, where 0.5 indicates
no predictive value and 1.0 indicates perfect performance.14
Because our dataset involved patients with multiple op-
erations, independence of observations could not be ex-
cluded beforehand. Therefore, to examine this phenomenon,
all regression analyses were first performed using conven-
tional techniques and repeated using generalized estimation
equations,15with “patient” as the classification factor. No
relevant differences were observed between the parameter
estimates as determined according to both methodologies.
Therefore, we concluded that interobservation correlation
was not a major issue in our dataset, and we present only the
results based on classical methods.
Results
A total of 52 387 surgical procedures were performed in
men, including 12 378 orthopedic surgeries (24%); 9273
ear, nose, and throat surgeries (18%); and 8637 abdominal
surgeries (16%). Among the 56 206 procedures in women,
gynecological surgery was most common with 15 312 pro-
cedures (27%), followed by orthopedic surgery with 9840
(18%), and abdominal surgery with 7816 (14%). Because of
reallocation of patients among regional hospitals, the annual
volumes of ophthalmic and gynecological procedures de-
creased in the early 1990s. During the study period, the
volume of orthopedic surgery gradually increased, whereas
the volume of abdominal surgery decreased slightly.
With a median age of 51 years (25th to 75th percentile:
34-65 years), men were 7 years older than were women
(median 44 years; 25th to 75th percentile: 32-62 years). The
majority of patients had a score of 0 points on the adapted
Lee index (Figure 1). During the study period, no systematic
change in scores on the adapted Lee index was observed in
men. By comparison, scores in women worsened slightly in
1136 The American Journal of Medicine, Vol 118, No 10, October 2005

Page 4

1993 and remained constant thereafter, a shift that was
strongly related to the decline in the number of gynecolog-
ical procedures.
A total of 1877 patients (1.7%) died perioperatively. A
cardiovascular complication was the principal (405 patients)
or the secondary (138 patients) cause of death in 543 pa-
tients (0.5% of the sample; 29% of deaths). Patients in
whom an autopsy report was available (326 patients; 17% of
deaths) were more often labeled as having cardiovascular
death than patients in whom no such report was available
(37% [122/326] vs 27% [421/1551] patients; P ?0.001).
Infection, the most common noncardiovascular cause of
death, was the primary cause in 231 patients and the sec-
ondary cause in 308 patients, representing 29% of deaths.
All-cause mortality and cardiovascular mortality were
higher in men than in women: 2.2% (1167/52 387) versus
1.3% (710/56 206) (P ?0.001), and 0.7% (350/52 387)
versus 0.3% (193/56 206) (P ?0.001). During the study
period, no systematic change in all-cause mortality was
observed in men (Figure 2). In contrast, all-cause mortality
in women increased significantly from 0.9% (55/6280) in
1991 to 1.3% (69/5151) in 1993 and 1.5% (79/5315) in
2000 (71% increase; P ?0.001). There were no significant
changes in cardiovascular mortality over time in either men
or women.
Important differences in the incidence of perioperative car-
diovascular death were observed in relation to type of surgery
(Table 1). Patients undergoing vascular surgery, especially
thoseundergoingaorticsurgery,hadthehighestcardiovascular
mortality (1.8%), followed by patients undergoing neurologic
surgery (1.7%), renal transplant (about 1.1%), and pulmonary
surgery (1.1%). Breast, dental, eye, and gynecology surgery
were associated with cardiovascular mortality rates below
0.1%. The 15 318 patients (14%) who had laparoscopic pro-
cedure had a lower incidence of cardiovascular death than did
patients who had open surgery (0.2% vs 0.6%; P ?0.001)
(Table 2). The 774 patients (0.7%) who underwent emergency
surgery had a significantly higher rate of cardiovascular death
than did patients who underwent nonemergency surgery (6.1%
vs 0.5%; P ?0.001).
In univariable analyses, the adapted Lee index and its
individual components were associated with an increased
risk of cardiovascular death (Table 2). The prospective C
statistic for the prediction of cardiovascular mortality in this
validation analysis was 0.63. The C statistic for this explor-
atory analysis was substantially higher in the subset of
66 530 low- to intermediate-risk surgical procedures (in-
cluding breast, carotid, dental, endocrine, eye, gynecology,
orthopedic, reconstructive, and urologic surgery) than in the
remaining 42 063 intermediate- to high-risk procedures
(0.68 vs 0.56). When more detailed information, including
the type of surgery (defined as low, low-intermediate, in-
termediate-high, and high), whether it was laparoscopic or
open, and whether it was emergent, was added to the Lee
index, the retrospective C statistic was 0.79 (Table 3). If age
were included, the C statistic rose further, to 0.85.
Figure 1
the bar), 2 points (grey portion), and ?3 points according to the cardiovascular risk index as developed by Lee and colleagues.8
Time trends in the cardiovascular risk profile. Data indicate the percentage of patients with a score of 1 point (white portion of
Figure 2
mortality (open circles) and cardiovascular death (closed circles).
Time trends in the incidence of perioperative all-cause
1137Boersma et al Prediction of perioperative cardiovascular mortality

Page 5

Discussion
Cardiovascular mortality still is a major burden in pa-
tients undergoing noncardiac surgery. In the investigated
cohort, about 7 of every 1000 procedures in men and 3 of
every 1000 procedures in women resulted in fatal in-
hospital cardiovascular complications. In contrast, anes-
thesia-related mortality occurs only in approximately 1 of
250 000 procedures.16Interestingly, patients who under-
went postmortem examination were considerably more
often classified as cardiovascular death than were pa-
tients in whom no such examination was performed,
suggesting that the incidence and effect of cardiovascular
complications after noncardiac surgery may be underes-
timated in clinical practice.
The Lee index (or revised Goldman index) is consid-
ered the best currently available cardiac risk prediction
index in noncardiac surgery because it was developed on
contemporary, prospectively gathered clinical data from
unselected patients who underwent a wide spectrum of
procedures and were followed systematically postopera-
tively, including standardized visits and cardiac biomar-
kers, for a range of clinically relevant cardiac out-
comes.17,18In our study, the Lee index was adapted for
an administrative database and its use extended to a
different goal––the use of administrative data specifically
to predict perioperative cardiovascular mortality. How-
ever, in agreement with another investigation,19our data
also suggest that the Lee index is probably suboptimal for
identifying patients with greater cardiac risk, perhaps
because it excluded emergency operations and perhaps
because the type of surgery, which is one of the main
determinants of adverse cardiovascular outcome,13was
considered in only 2 subtypes: high risk, including intra-
peritoneal, intrathoracic and suprainguinal vascular pro-
cedures; and all remaining nonlaparascopic procedures,
mainly including orthopedic, abdominal, and other vas-
cular procedures. We found that a more subtle classifi-
cation, as suggested by the American Heart Association/
American College of Cardiology guideline committee,13
resulted, at least retrospectively, in a substantially better
risk discrimination. We realize that the Lee index was
developed for the prediction of prospectively detected
“major cardiac complications” (which included myocar-
dial infarction, cardiogenic pulmonary edema, cardiac
arrest, and cardiac death) and not for the prediction of
cardiovascular death only. It is unknown whether the C
statistics would have been more favorable if we had used
the same endpoint as Lee and colleagues, but some be-
lieve it may be easier to predict the incidence of death
than to predict a broader range of clinical outcomes.20In
addition, from the perspective of assessing quality of care
Table 1
Perioperative cardiovascular and all-cause death in patients undergoing noncardiac surgery for various indications
Type of surgery
Number of
procedures
Cardiovascular death
All-cause death
Primary
cause
Secondary
causeTotal
Number (%)
Abdominal
Hepatico, pancreatico, billiary
Esophagogastric
Other abdominal
Breast
Dental
Ear, nose, throat
Endocrine
Eye
Gynecology
Neurologic
Orthopedic
Reconstructive
Pulmonary
Renal transplant
Urologic
Vascular
Aortic-acute
Aortic-elective
Carotid endarterectomy
Peripheral bypass
Other vascular
Other
Any type
16 453
2752
11 982
3714
2411
1225
15 291
1029
9163
15 343
5797
22 218
4157
1965
711
11 116
6234
196
890
891
927
3854
9423
108 593
63 (0.4)
10 (0.4)
53 (0.4)
12 (0.3)
0 (0.0)
1 (0.1)
114 (0.7)
1 (0.1)
1 (0.0)
2 (0.0)
87 (1.5)
34 (0.2)
5 (0.1)
14 (0.7)
8 (1.1)
28 (0.3)
90 (1.4)
21 (10.7)
29 (3.3)
4 (0.4)
14 (1.5)
36 (0.9)
18 (0.2)
405 (0.4)
50 (0.3)
4 (0.1)
44 (0.4)
7 (0.2)
0 (0.0)
0 (0.0)
22 (0.1)
2 (0.2)
0 (0.0)
1 (0.0)
14 (0.2)
9 (0.0)
2 (0.0)
7 (0.4)
0 (0.0)
10 (0.1)
25 (0.4)
7 (3.6)
7 (0.8)
2 (0.2)
2 (0.2)
13 (0.3)
13 (0.1)
138 (0.1)
113 (0.7)
14 (0.5)
97 (0.8)
19 (0.5)
0 (0.0)
1 (0.1)
136 (0.9)
3 (0.3)
1 (0.0)
3 (0.0)
101 (1.7)
43 (0.2)
7 (0.2)
21 (1.1)
8 (1.1)
38 (0.3)
115 (1.8)
28 (14.3)
36 (4.0)
6 (0.7)
16 (1.7)
49 (1.3)
31 (0.3)
543 (0.5)
606 (3.7)
129 (4.7)
488 (4.1)
122 (3.3)
0 (0.0)
2 (0.2)
411 (2.7)
6 (0.6)
11 (0.1)
20 (0.1)
381 (6.6)
116 (0.5)
12 (0.3)
86 (4.4)
14 (2.0)
159 (1.4)
277 (4.4)
57 (29.1)
72 (8.1)
18 (2.0)
28 (3.0)
142 (3.7)
92 (1.0)
1877 (1.7)
1138 The American Journal of Medicine, Vol 118, No 10, October 2005

Page 6

using administrative databases, cardiovascular mortality
is certainly a relevant endpoint.
The American Heart Association/American College of
Cardiology guidelines identify advanced age as a minor
predictor of cardiovascular risk.13In our study, as in essen-
tially every study of perioperative risk, cardiovascular mor-
tality increased progressively with age.3-11Indeed, elderly
patients might often have asymptomatic coronary disease,
which places them at increased risk of perioperative cardio-
vascular complications. The key question has been whether
age itself is an independent predictor or whether its impor-
tance is subsumed by its strong relation with other measur-
able evidence of the severity of disease or of comorbid
conditions. Our finding of the major importance of age per
se might be because our approach to assessing risk factors
was by means of the medical history as coded according to
the ICD-9 system and subsequently entered in an electronic
database by administrative personnel based on written in-
formation provided by health-care professionals. These em-
ployees are specifically instructed to avoid inappropriate
over-diagnosis; as a result, key medical conditions might
have been overlooked, and, consequently, the relative con-
tribution of these factors to cardiovascular death might have
been underestimated, thereby also increasing the apparent
Table 2
death
Univariable relation among the Lee index, demographic and clinical characteristics, and perioperative cardiovascular
Characteristic
Number of
procedures
Number (%)
cardiovascular death Crude, unadjusted odds
ratio
(95% confidence interval)
Number (%)
Adaptation of the Lee index score
and its components (see
Methods)
Adapted Lee index
?3
2
1
0
High risk
Not high risk
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
969
3380
28 892
75 352
29 426
79 167
3588
105 005
1377
107 216
35 (3.6)
57 (1.7)
196 (0.7)
255 (0.3)
224 (0.8)
319 (0.4)
77 (2.1)
466 (0.4)
50 (3.6)
493 (0.5)
11 (2.2)
532 (0.5)
31 (1.6)
512 (0.5)
36 (1.8)
507 (0.5)
11.0 (7.7-15.8)
5.1 (3.8-6.7)
2.0 (1.7-2.4)
1
1.9 (1.6-2.3)
1
4.9 (3.9-6.3)
1
8.2 (6.1-11.0)
1
4.6 (2.5-8.3)
1
3.5 (2.4-5.0)
1
3.8 (2.7-5.4)
1
Type of surgery*
History of ischemic heart disease
History of heart failure
History of CVA 500
108 093
1894
106 699
2001
106 592
Renal insufficiency
Diabetes mellitus
Detailed data on type of surgery
Type of surgery†High risk
Intermediate-high risk
Low-intermediate risk
Low risk
Yes
No
Yes
No
1078
40 985
33 275
33 255
15 318
93 275
774
107 819
63 (5.8)
371 (0.9)
81 (0.2)
28 (0.1)
24 (0.2)
519 (0.6)
47 (6.1)
496 (0.5)
73.6 (46.9-115)
10.8 (7.4-15.9)
2.9 (1.9-4.4)
1
0.3 (0.2-0.4)
1
14.0 (10.2-19.0)
1
Laparascopic procedure
Emergency surgery
Other potential risk determinants
Age (years)
?80
70-80
60-70
50-60
40-50
?40
5314
12 619
15 742
15 675
16 987
42 256
77 (1.5)
165 (1.3)
146 (0.9)
91 (0.6)
37 (0.2)
27 (0.1)
23.0 (14.8-35.7)
20.7 (13.7-31.1)
14.6 (9.7-22.1)
9.1 (5.9-14.0)
3.4 (2.1-5.6)
1
CVA ? cerebrovascular accident.
*According to the Lee index: high risk ? intraperitoneal, intrathoracic, and suprainguinal vascular procedures; not high risk ? other procedures.
†According to the Amercian Heart Association/American College of Cardiology classification: high risk ? aortic; intermediate risk ? abdominal; ear,
nose, throat; neurologic; orthopedic; pulmonary; renal transplant; urologic; vascular, excluding aortic and carotid; low risk ? breast; carotid; dental;
endocrine; eye; gynecology; reconstructive. Within the intermediate risk group, patients undergoing orthopedic or urologic surgery had significantly
lower risk than patients undergoing other types of surgery. Therefore, we labeled orthopedic and urologic procedures as low-intermediate risk, and the
remaining procedures as intermediate-high risk.
1139Boersma et al Prediction of perioperative cardiovascular mortality

Page 7

independent contribution of age itself.21In addition, we
restricted our analyses to patients who underwent surgery.
No information was included from patients who were
screened but who did not undergo surgery because the risk
was perceived as prohibitive. Obviously, exclusion of pa-
tients at risk of adverse cardiovascular outcome might have
diluted estimates of relative risk.
The identification of patients at risk of perioperative
cardiovascular complications has improved considerably
in recent years. Beta-blockers reduce complication rates
for some categories of patients, such as those undergoing
major vascular surgery,22-24and statins may be useful as
well.25By comparison, routine coronary revasculariza-
tion is not beneficial.26The development and implemen-
tation of such strategies for the entire spectrum of surgi-
cal patientsremains an
contemporary medicine.27In that regard, it is noteworthy
that the incidence of fatal perioperative cardiovascular
complications at our center did not decline during the
10-year study period.
importantchallengefor
Conclusion
This single center study, which involved over 100 000
subjects, demonstrated that perioperative cardiovascular
mortality is a major burden in patients undergoing non-
cardiac surgery. Little progress has been achieved in
reducing cardiovascular mortality during the years of the
study. The adapted Lee index had an admirable perfor-
mance to predict cardiovascular mortality, but its simple
classification of procedures as high risk versus not high
risk seems suboptimal. Our analysis is limited by the fact
that our data are derived from an administrative database,
the retrospective nature of the data, the ICD-9 coding of
clinical characteristics, and our evaluation of cardiovas-
cular death rather than a broader range of clinical com-
plications. Our approach seems most applicable to situ-
ations in which administrative data are to be used to
assess outcomes and to compare outcomes in different
hospitals or regions. Furthermore, prospective studies are
warranted to confirm our findings.
Table 3
analyses of 108 593 subjects undergoing noncardiac surgery
Multivariable relation among the Lee index, type of surgery, age, and perioperative cardiovascular death, based on
Characteristic
Odds ratio (95% confidence interval)
Adapted Lee index
only
Adapted Lee index
and type of
surgery
Adapted Lee index,
type of surgery
and age
Model C statistic
Adapted Lee index
0.63
11.0 (7.7-15.8)
5.1 (3.8-6.7)
2.0 (1.7-2.4)
1
0.790.85
?3
2
1
0
?3 Adapted Lee index, excluding type of
surgery*
9.2 (5.5-15.4)6.4 (3.8-10.8)
2
1
0
High risk
Intermediate-high risk
Low-intermediate risk
Low risk
Yes
No
Yes
No
?80
70-80
60-70
50-60
40-50
?40
5.6 (4.0-7.9)
2.3 (1.8-3.0)
1
35.7 (22.1-57.6)
10.3 (7.0-15.2)
2.8 (1.8-4.3)
1
0.3 (0.2-0.4)
1
4.6 (3.2-6.5)
1
4.0 (2.9-5.6)
1.7 (1.3-2.2)
1
20.0 (12.3-32.5)
10.3 (7.0-15.1)
2.7 (1.7-4.1)
1
0.3 (0.2-0.4)
1
4.4 (3.1-6.4)
1
19.9 (12.8-31.1)
12.6 (8.3-19.0)
8.5 (5.6-12.9)
5.6 (3.6-8.7)
2.5 (1.5-4.1)
1
Type of surgery†
Laparascopic procedure
Emergency surgery
Age (years)
*Index that assigns one point to each of the following characteristics: ischemic heart disease, history of heart failure, history of cerebrovascular
disease, renal insufficiency, and diabetes mellitus.
†According to the Amercian Heart Association/American College of Cardiology classification: high risk ? aortic; intermediate risk ? abdominal; ear,
nose, throat; neurologic; orthopedic; pulmonary; renal transplant; urologic; vascular, excluding aortic and carotid; low risk ? breast; carotid; dental;
endocrine; eye; gynecology; reconstructive. Within the intermediate risk group, patients undergoing orthopedic or urologic surgery had significantly
lower risk than patients undergoing other types of surgery. Therefore, we labeled orthopedic and urologic procedures as low-intermediate risk, and the
remaining procedures as intermediate-high risk.
1140The American Journal of Medicine, Vol 118, No 10, October 2005

Page 8

References
1. Heart Protection Study Collaborative Group. MRC/BHF Heart protec-
tion study of cholesterol lowering with simvastatin in 20536 high-risk
individuals: a randomised placebo controlled trial. Lancet. 2002;360:
7–22.
2. Dawood MM, Gutpa DK, Southern J, et al. Pathology of fatal periop-
erative myocardial infarction: implications regarding pathophysiology
and prevention. Int J Cardiol. 1996;57:37–44.
3. Goldman L, Caldera DL, Nussbaum SR, et al. Multifactorial index of
cardiac risk in noncardiac surgical procedures. N Engl J Med. 1997;
297:845–850.
4. Zeldin RA. Assessing cardiac risk in patients who undergo noncardiac
surgical procedures. Can J Surg. 1984;27:402–404.
5. Larsen SF, Olesen KH, Jacobsen E, et al. Prediction of cardiac risk in
non-cardiac surgery. Eur Heart J. 1987;8:179–185.
6. Ashton CM, Petersen NJ, Wray NP, et al. The incidence of perioper-
ative myocardial infarction in men undergoing noncardiac surgery.
Ann Intern Med. 1993;118:504–510.
7. Detsky AS, Abrams HB, McLaughlin JR, et al. Predicting cardiac
complications in patients undergoing non-cardiac surgery. J Gen In-
tern Med. 1986;1:211–219.
8. Lee TH, Marcantonio ER, Mangione CM, et al. Derivation and pro-
spective validation of a simple index for prediction of cardiac risk of
major noncardiac surgery. Circulation. 1999;100:1043–1049.
9. Mangano DT, Goldman L. Preoperative assessment of patients with
known or suspected coronary disease. N Engl J Med. 1995;333:1750–
1756.
10. Eagle KA, Coley CM, Newell JB, et al. Combining clinical and
thallium data optimizes preoperative assessment of cardiac risk before
major vascular surgery. Ann Intern Med. 1989;110:859–866.
11. Boersma E, Poldermans D, Bax JJ, et al. Predictors of cardiac events
after major vascular surgery: Role of clinical characteristics, dobut-
amine echocardiography, and beta-blocker therapy. JAMA. 2001;285:
1865–1873.
12. ICD-9-CM: 2002:International Classification of Diseases, 9th Revi-
sion, Volumes 1 and 2. American Medical Association, 2001.
13. Eagle KA, Berger PB, Calkins H, et al. American College of Cardi-
ology/American Heart Association Task Force on Practice Guidelines
(Committee to Update the 1996 Guidelines on Perioperative Cardio-
vascular Evaluation for Noncardiac Surgery). ACC/AHA guideline
update for perioperative cardiovascular evaluation for noncardiac sur-
gery––executive summary report of the American College of Cardi-
ology/American Heart Association Task Force on Practice Guidelines
(Committee to Update the 1996 Guidelines on Perioperative Cardio-
vascular
2002;105:1257–1267.
14. Kendal DG. Rank Correlation Methods. London, England: Charles
Griffin; 1962.
15. Liang KY, Zeger SL. Longitudinal data analysis using general linear
models. Biometrika. 1986;73:13–22.
16. Eichhorn JH. Prevention of intraoperative anesthesia accidents and
related severe injury through safety monitoring. Anesthesiology. 1989;
70:572–577.
17. Lindenauer PK, Pekow P, Wang K, et al. Lipid-lowering therapy and
in-hospital mortality following major noncardiac surgery. JAMA.
2004;291:2092–2099.
18. Grayburn PA, Hillis LD. Cardiac events in patients undergoing non-
cardiac surgery: shifting the paradigm from noninvasive risk stratifi-
cation to therapy. Ann Intern Med. 2003;138:506–511.
19. Ridley S. Cardiac scoring systems–what is their value? Anaesthesia.
2003;58:985–991.
20. Boersma E, Pieper KS, Steyerberg EW, et al. Predictors of outcome in
patients with acute coronary syndromes without persistent ST-segment
elevation: results from an international trial of 9461 patients. Circu-
lation. 2000;101:2557–2567.
21. Iezzoni LI. Using administrative diagnostic data to assess the quality
of hospital care. Pitfalls and potential of ICD-9-CM. Int J Technol
Assess Health Care. 1990;6:272–281.
22. Mangano DT, Layug EL, Wallace A, Tateo I. Effect of atenolol on
mortality and cardiovascular morbidity after noncardiac surgery. Mul-
ticenter study of the perioperative ischemia research group. N Engl
J Med. 1997;335:1713–1720.
23. Poldermans D, Boersma E, Bax JJ, et al. The effect of bisoprolol on
perioperative mortality and myocardial infarction in high-risk patients
undergoing vascular surgery. N Engl J Med. 1999;341:1789–1794.
24. Poldermans D, Bax JJ, Kertai MD, et al. Statins are associated with a
reduced incidence of perioperative mortality in patients undergoing
major noncardiac vascular surgery. Circulation. 2003;107:1848–1851.
25. Lindenauer PK, Pekow P, Wang KJ, et al. Lipid-lowering therapy and
in-hospital mortality following major noncardiac surgery. JAMA.
2004;291:2092–2099.
26. McFalls EO, Ward HB, Moritz TE, et al. Coronary-artery revascular-
ization before elective major vascular surgery. N Engl J Med. 2004;
351:2795–2804.
27. Schouten O, Poldermans D, Visser L, et al. Fluvastatin and bisoprolol
for the reduction of perioperative cardiac mortality and morbidity in
high-risk patients undergoing non-cardiac surgery: rationale and de-
sign DECREASE-IV study. Am Heart J. 2004;148:1047–1052.
Evaluationfor NoncardiacSurgery).
Circulation.
1141 Boersma et al Prediction of perioperative cardiovascular mortality