Preoperative risk prediction of surgical site infection

requiring hospitalization or reoperation in patients

undergoing vascular surgery
Surbhi Leekha, MBBS, MPH,a Brian D. Lahr, MS,b Rodney L. Thompson, MD,c
Priya Sampathkumar, MD,c Audra A. Duncan, MD,d and Robert Orenstein, DO,e Baltimore, Md;
Rochester, Minn; and Scottsdale, Ariz

Objective: The objective of this study was to develop a surgical site infection (SSI) prediction score for risk assessment
before elective vascular surgery.
Methods: We conducted a nested case-control study among patients who underwent elective vascular (abdominal aortic and
peripheral arterial) surgery from January 1, 2003, to December 31, 2007, at Mayo Clinic (Rochester, Minn) an academic
tertiary surgical center. Cases were patients with SSI requiring hospitalization; controls (one or two per case) were
matched on type of procedure and date of surgery. Clinical data were collected by chart review. A risk score based on
preoperative variables was developed using multivariable logistic regression and bootstrap resampling. The C statistic,
equivalent to the area under the receiver operating characteristic curve, was used to assess discrimination. Calibration was
assessed by plotting percentile risk groups of model-predicted values against observed proportions of subjects with SSI.
Results: Eighty-four cases were compared with 160 controls. Preoperative variables independently associated with SSI risk
were critical limb ischemia, previous SSI, prior revascularization procedure, and chronic obstructive pulmonary disease.
A prediction model containing these variables was developed (model and risk score C statistic of 0.737 and 0.727,
respectively). The calibration curve did not appear to deviate appreciably from the 45-degree line of identity.
Conclusions: We developed an SSI risk score based on noninvasive preoperative variables with acceptable discrimination
and calibration. This tool needs prospective and external validation. (J Vasc Surg 2016;64:177-84.)

A variety of patient, procedural, and provider-related
factors contribute to the development of surgical site infec-
tion (SSI). Using these risk factors, investigators have
attempted to develop scoring systems to help predict and
to stratify patients’ risk of SSI after surgery.1-11 The
primary limitation of these scoring systems (including the
recently modified Centers for Disease Control and Preven-
tion’s National Healthcare Safety Network [NHSN] risk
models10) is that most are based on combinations of
preoperative, intraoperative, and postoperative factors.
Because of their reliance on variables that can be assessed
only during or after surgery, such tools do not help with
prediction and potential modification of infection risk
From the Department of Epidemiology and Public Health, University of
Maryland School of Medicine, Baltimorea; the Division of Biomedical
Statistics and Informatics,b Division of Infectious Diseases,c and Division
of Vascular and Endovascular Surgery,d Mayo Clinic, Rochester; and the
Division of Infectious Diseases, Mayo Clinic, Scottsdale.e
This study was supported by the Small Grants Program at Mayo Clinic,
Rochester, Minn.
Author conflict of interest: none.
Presented in a poster at the Fifth Decennial International Conference on
Healthcare-Associated Infections, Atlanta, Ga, March 18-22, 2010.
Correspondence: Surbhi Leekha, MBBS, MPH, 110 S Paca St, 6th Fl, Bal-
timore, MD 21201 (e-mail: sleekha@epi.umaryland.edu).
The editors and reviewers of this article have no relevant financial relationships
to disclose per the JVS policy that requires reviewers to decline review of any
manuscript for which they may have a conflict of interest.
0741-5214
Copyright Ó 2016 by the Society for Vascular Surgery. Published by
Elsevier Inc.
http://dx.doi.org/10.1016/j.jvs.2016.01.029
before surgery. Only a few studies report scoring systems
based on preoperative variables alone for SSI risk prediction
after major surgery.2,7,11,12
Despite advances in SSI prevention, the rates of SSI af-
ter vascular surgical procedures remain high. Based on the
2009 NHSN surveillance data, the median SSI rate after
peripheral vascular bypass in the United States was 4.6
(interquartile range, 2.8-8.5) per 100 surgeries. Therefore,
a preoperative SSI risk stratification system could be useful
in vascular surgery, in which SSI rates are significantly
higher than in other “clean” procedures, infection can
result in loss of limb or life, and nonoperative options exist
for patients deemed to be at high SSI risk.
We undertook this study to develop a prediction score
based on preoperative factors that could serve as a simple
SSI risk assessment tool before elective vascular surgery.
METHODS
Study design and setting. We conducted a nested
case-control study among patients undergoing elective
vascular surgery involving the abdominal aorta or periph-
eral arteries (including endovascular procedures requiring
surgical incisions but excluding thoracic aortic, upper limb,
and carotid procedures and percutaneous endovascular
procedures) during a 5-year period between January 1,
2003, and December 31, 2007, at Mayo Clinic in
Rochester, Minnesota. Mayo Clinic is a tertiary care
medical center with a high vascular surgery volume
(>800 cases per year) and an active infection prevention
program that oversees routine surveillance for SSI after all

177

178 Leekha et al
vascular surgical procedures. No significant changes in
surveillance methods occurred during the study period.
The study was approved by the Mayo Clinic Institutional
Review Board as a minimal risk protocol with waiver of
Health Insurance Portability and Accountability Act
authorization and informed consent.
Case and control selection. All patients who under-
went vascular surgery involving the abdominal aorta or pe-
ripheral arteries (excluding thoracic aortic, upper limb, and
carotid procedures) during the study period were identified
through the Mayo Clinic Surgical Information Recording
System that routinely captures data on all surgical cases in
both the inpatient and outpatient settings in a prospective
manner. All patients are seen by the surgeon in follow-up
within the first few weeks of surgery, and most patients
with infections would return to Mayo Clinic for manage-
ment of complications. Cases of SSI were identified using
data collected by trained infection preventionists for
routine SSI surveillance. This surveillance is performed by
monitoring all hospital admissions, surgical case listings,
microbiology reports, and infectious diseases consultations,
using NHSN criteria to define infections.13 Postdischarge
questionnaires were not routinely sent during the study
period, but the surgeon or the infection prevention and
control department would usually be notified of hospitali-
zations at other facilities. The time period for SSI surveil-
lance was up to 30 days for surgeries without implant of
foreign material and up to 1 year for surgeries in which
foreign material was implanted, per NHSN criteria.13 We
included all deep and organ/space SSIs as cases, consistent
with the recommendations of the National Quality Forum
for public reporting, because these infections often require
hospitalization and are associated with significant morbidity
and mortality, in comparison with superficial infections,
which are often treated in outpatient settings, have low
inter-rater reliability, and are difficult to validate.14,15
However, to capture more serious superficial SSIs, we
also included those superficial SSIs that required hospital
stay for intravenous antibiotics and wound care or operative
intervention during the index admission or readmission.
From the remaining cohort of patients who did not develop
SSI during the surveillance period, we individually matched
two controls per case, based on date of surgery (62 weeks)
and type of procedure (in three broad categories of pro-
cedures involving the abdominal aorta, peripheral vascular
bypass, and nonbypass lower limb arterial procedures).
Patients who had previously denied research authorization
were excluded from both the case and control groups, in
accordance with Minnesota state law.
Exposure assessment. We performed detailed chart re-
view to collect information on several preoperative variables
to evaluate their association with SSI and for assessment as
candidates for inclusion in the prediction model (Table I).
Statistical analyses. Descriptive statistics on pre-
operative variables are presented as median (range or
interquartile range), mean (standard deviation), or count
(percentage) as appropriate. All candidate risk factors were
tested individually for an association with SSI case status
JOURNAL OF VASCULAR SURGERY
July 2016
using univariable logistic regression, with only those
detected at an a level of .1 carried forward in multivariable
analysis. These screened variables were then entered in a
multivariable logistic regression model, which was reduced
to the most important predictors using stepwise variable
selection with backwards elimination (a level of .05). Both
regular logistic and conditional logistic regression were
used and yielded very similar results; therefore, uncondi-
tional logistic regression was used to minimize loss of infor-
mation due to missing data. We performed tests of both
discrimination and calibration to assess the predictive accu-
racy of the model. Discrimination was assessed using the C
statistic, which is the area under the receiver operating char-
acteristic (ROC) curve, a plot of sensitivity vs 1
specificity
of the model; a value of 0.5 indicates random predictions,
and a value of 1.0 indicates perfect predictions. To internally
validate the fit and performance of the model, bootstrap
resampling was used. For each of 400 bootstrap resamples
(selected through random sampling with replacement from
the original set of patients, each of an equal sample size),
the modeling procedure was repeated with the same stepwise
criteria so that variability in the automated variable selection
could be assessed. By repeatedly fitting the model selection
in a bootstrap resample and evaluating the performance of
the model on the original sample (in terms of the model C
statistic), bias due to overfitting was estimated as the average
difference in the performance between the bootstrap and test
models across all resamples. A bias-corrected C statistic was
then obtained by subtracting the optimism bias from the
index derived on the original sample. Furthermore, the
frequency of “selected” variables was summarized as a
percentage across all bootstrap resamples, and only
predictors consistently retained in the modeling ($60% of
samples) were included in the risk score.
To assess calibration, which is the model’s ability to
predict accurately the absolute level of risk that is subse-
quently observed, the matched cases and controls were
weighted inversely to their original sampling frame, thereby
rescaling the model predictions to the entire cohort at risk.
The accuracy of model-predicted values relative to
observed values was assessed visually by plotting percentile
risk groups against observed proportions of subjects with
SSI, with a nonlinear calibration curve using the nonpara-
metric loess smooth estimator.
From the final selected model, regression coefficients
were used to derive a preoperative risk score for predicting
SSI. In particular, points were assigned for the presence of
each risk factor in the model, weighted according to the
respective regression coefficient rounded to the nearest
integer. Using these point values, a subject’s risk score is
simply computed as the aggregate number of points from
the risk factor profile, with higher scores corresponding
to increased risk of SSI.
RESULTS
Between January 1, 2003, and December 31, 2007, a
total of 4302 surgical procedures involving the abdominal
aorta and peripheral arteries (excluding thoracic aortic,
JOURNAL OF VASCULAR SURGERY
Volume 64, Number 1 Leekha et al 179

Table I. Distribution of candidate preoperative variables for vascular surgical site infection (SSI) risk prediction models
between cases and controls

No. with Univariable result

Variable missing data Cases (n ¼ 84) Controls (n ¼ 160) OR (95% CI) P valuea

Age 68.8 6 11.8 68.6 6 11.3 1.01 (0.80-1.27) .936

Female gender 30 (36) 51 (32) 1.19 (0.68-2.07) .545
Body mass index 1
<20 5 (6) 9 (6) 1.44 (0.42-4.91) F test, P ¼ .702
20-25 17 (20) 44 (28) 1.0 (ref)
25-30 32 (38) 61 (38) 1.36 (0.67-2.75)
30þ 30 (36) 45 (28) 1.73 (0.84-3.57)
Smoking status
Never smoker 13 (15) 27 (17) 1.0 (ref) F test, P ¼ .711
Current smoker 29 (35) 47 (29) 1.28 (0.57-2.87)
Past smoker 42 (50) 86 (54) 1.01 (0.48-2.16)
Functional statusb
Difficulty preparing meals 132 5 (13) 9 (13) 1.00 (0.31-3.22) 1.000
Difficulty feeding 11 0 (0) 6 (4) 0.23 (0.00-1.22) .155
Difficulty dressing 11 5 (6) 10 (7) 0.95 (0.31-2.89) .933
Difficulty using toilet 11 2 (3) 5 (3) 0.76 (0.14-4.00) .745
Difficulty housekeeping 11 18 (23) 26 (17) 1.42 (0.72-2.78) .309
Difficulty climbing stairs 11 42 (53) 55 (36) 1.97 (1.14-3.41) .016
Difficulty bathing 11 8 (10) 9 (6) 1.78 (0.66-4.80) .256
Difficulty walking 11 35 (44) 55 (36) 1.39 (0.80-2.41) .246
Difficulty using 11 11 (14) 22 (14) 0.95 (0.44-2.07) .897
transportation
Difficulty getting in or out 129 1 (3) 1 (1) 1.90 (0.12-31.17) .654
of bed
Difficulty managing 12 6 (8) 11 (7) 1.06 (0.38-2.99) .910
medication
Use of assistive device for 4 44 (53) 53 (34) 2.21 (1.29-3.81) .004
ambulation
Congestive heart failure 5 (6) 10 (6) 0.95 (0.31-2.87) .927
Current aspirin use 51 (61) 110 (69) 0.70 (0.40-1.22) .209
Current warfarin use 21 (25) 33 (21) 1.28 (0.69-2.40) .435
Current statin use 45 (54) 95 (59) 0.79 (0.46-1.34) .384
Diabetes mellitus
No 48 (57) 99 (62) 1.0 (ref) F test, P ¼ .057
Yes, insulin requiring 19 (23) 28 (18) 1.40 (0.71-2.75)
Yes, on oral agents 14 (17) 14 (9) 2.06 (0.91-4.67)
Yes, diet controlled 3 (4) 19 (12) 0.33 (0.09-1.15)
Hyperlipidemia 59 (70) 115 (72) 0.92 (0.52-1.65) .788
Active or untreated 2 (2) 3 (2) 1.28 (0.21-7.79) .791
malignant disease
Recent chemotherapy 0 (0) 0 (0) d
Radiation therapy at surgical 2 (2) 4 (3) 0.95 (0.17-5.30) .955
site
COPD 27 (32) 29 (18) 2.14 (1.16-3.94) .014
On any oxygen at home? 4 (5) 3 (2) 2.62 (0.57-11.97) .215
Chronic liver disease 5 (6) 2 (1) 5.00 (0.95-26.35) .058
Chronic kidney disease (GFR <60 on $2 occasions over 3 months in last year or diagnosis in chart)
No 63 (75) 121 (76) 1.0 (ref) F test, P ¼ .411
Yes, not on dialysis 14 (17) 32 (20) 0.84 (0.42-1.69)
Yes, on dialysis 7 (8) 7 (4) 1.92 (0.65-5.72)
Immunosuppressive 7 (8) 11 (7) 1.23 (0.46-3.30) .679
medication including
corticosteroid use in last
6 months
Recent weight loss (>10 lb 2 11 (13) 13 (8) 1.72 (0.73-4.02) .214
in last 3 months or >10%
body weight within last
6 months)
Recent (in last month) 2 (2) 6 (4) 0.63 (0.12-3.17) .572
infection at another body
site
(Continued on next page)
 JOURNAL OF VASCULAR SURGERY
180 Leekha et al July 2016

Table I. Continued.

No. with Univariable result

Variable missing data Cases (n ¼ 84) Controls (n ¼ 160) OR (95% CI) P valuea

Recent (in last month) 7 (8) 14 (9) 0.95 (0.37-2.45) .913

limb gangrene
Skin ulcers at time of surgery 28 (33) 35 (22) 1.79 (0.99-3.22) .053
Chronic skin condition 4 (5) 12 (8) 0.62 (0.19-1.98) .416
(eg, psoriasis)
Critical limb ischemia 43 (51) 41 (26) 3.04 (1.75-5.31) <.001
Previous surgery at same site 36 (43) 50 (31) 1.65 (0.96-2.85) .072
Previous surgical or 66 (79) 90 (56) 2.85 (1.55-5.24) <.001
interventional radiology-
guided peripheral
revascularization
procedure
Previous SSI 28 13 (18) 5 (4) 5.84 (1.99-17.10) .001
ASA score
2 4 (5) 10 (6) 1.0 (ref) F test, P ¼ .221
3 70 (83) 141 (88) 1.24 (0.38-4.10)
4 10 (12) 9 (6) 2.78 (0.64-12.06)
Recent hospitalization 30 (36) 47 (29) 1.34 (0.76-2.34) .312
(within 3 months)
Residence in nursing home 3 (4) 3 (2) 1.94 (0.38-9.82) .424
before surgery
Fasting glucose morning of 70 123.0 (102.0-146.0) 120.0 (103.0-140.0) 1.29 (0.42-3.96) .651
surgeryc
Fasting glucose (highest 41 114.0 (93.0-143.0) 106.5 (93.0-138.0) 1.21 (0.50-2.92) .673
value) in last 3 monthsc
Highest serum creatinine in 86 1.2 (1.0-1.5) 1.2 (1.0-1.5) 0.92 (0.49-1.75) .807
last weekc
Highest serum creatinine in 1.2 (1.0-1.5) 1.2 (1.0-1.5) 1.10 (0.66-1.83) .716
last 3 monthsc
Highest WBC count in last 91 8.1 6 2.3 8.0 6 2.6 1.01 (0.88-1.15) .906
week

ASA, American Society of Anesthesiologists; CI, confidence interval; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; OR, odds
ratio; WBC, white blood cell.
Categorical variables are presented as number (%). Continuous variables are presented as mean 6 standard deviation or median (interquartile range).
a
Effects for all variables tested using logistic regression.
b
Effect measured from exact logistic regression because of sparse data.
c
Estimated effect from logistic regression based on log-transformed data.

carotid, and upper limb procedures) were performed at our
institution. Of these, 150 (3.5%) resulted in SSI based on
NHSN criteria (32 organ/space infections, 19 deep inci-
sional, and 99 superficial incisional infections). Of these,
84 infections met criteria for inclusion as cases in this study
and were compared with 160 matched controls. The distri-
bution of potential SSI risk predictor variables between
cases and controls is summarized in Table I.
SSI risk scores. Using multivariable logistic regression
and stepwise variable selection with backwards elimination,
the following preoperative variables were consistently
selected across the bootstrap resamples ($60%) and thus
deemed robust predictors of SSI: critical limb ischemia,
previous SSI, prior revascularization procedure, and
chronic obstructive pulmonary disease (COPD). The C
statistic for the selected preoperative model based on the
original fit was 0.737, which represents fairly good discrim-
ination (Table II). However, the estimated optimism bias
due to variability in the automated model selection was
0.047, and thus a bias-corrected estimate of 0.690 would
provide a reasonable approximation of the model C statistic
in a future independent validation. Using a smoothing
estimator relating observed outcomes to predicted values
(rescaled to convey absolute risk of SSI), the calibration
curve of the preoperative model did not appear to deviate
appreciably from the 45-degree line of identity (ie, ideal
calibration), suggesting acceptable calibration (Fig 1).
On the basis of the final multivariable prediction
model, regression coefficients were used to derive a preop-
erative risk score for development of SSI after vascular
surgery. Points were assigned for the presence of each
risk factor, weighted in magnitude by the corresponding
regression coefficients rounded to the nearest integer,
and summed together to define an individual’s risk score.
For example, a patient with COPD and critical limb
ischemia but no previous revascularization procedure or
history of SSI would have a preoperative risk score of
2 of 5 possible points. From a logistic model with risk score
included as a continuous variable, an ROC curve was
derived and superimposed over the ROC curve from
the corresponding multivariable model for comparison.
In Fig 2, the ROC curves from the multivariable prediction
JOURNAL OF VASCULAR SURGERY
Volume 64, Number 1 Leekha et al 181

Table II. Preoperative risk factors for surgical site

infection (SSI) after vascular surgery and development of
preoperative SSI risk score

Multivariable result
OR (95% CI) Cb
Variable [P value] b (C ¼ 1) Points

Critical ischemia 2.91 (1.61-5.27) 1.07 1.07 1

[<.001]
COPD 2.10 (1.07-4.09) 0.74 0.74 1
[.030]
Prior revascularization 2.68 (1.38-5.22) 0.99 0.99 1
procedure [.004]
Previous SSI
Yes 6.56 (2.08-20.65) 1.88 1.88 2
[.001]
No 1.0 (reference) d d 0
No previous surgery 1.63 (0.63-4.24) 0.49 0.49 0
[.313]
Model C statistic 0.737
Bias-corrected 0.690
C statistic

CI, Confidence interval; COPD, chronic obstructive pulmonary disease;

OR, odds ratio. Fig 1. Calibration plot of the preoperative multivariable predic-
tion model. SSI, Surgical site infection.

model and risk score models are similar (C statistic, or area

under the ROC curve, of 0.737 vs 0.727, respectively),
showing that the risk score adequately summarizes the
multivariable model. The bubble plot in Fig 3 displays
the range of preoperative scores in relation to absolute
risk of SSI, with the size of the bubble proportional to
the frequency observed in the analysis set.

DISCUSSION
In this study, we have developed a simple SSI risk strat-
ification model with acceptable discrimination and calibra-
tion for use before elective vascular surgery. This
preoperative score has the distinct advantage of providing
risk prediction before elective surgery.
The variables in our prediction model, namely, critical
limb ischemia, prior revascularization procedure, COPD,
and previous SSI, are easily identified in patients, and the
resulting risk score can be easily calculated during an office
visit. Further, these variables have biologic plausibility as
risk factors for SSI. Both COPD and critical limb ischemia
are associated with tissue hypoxia and have previously been
identified as risk factors for SSI after lower extremity vascular
Fig 2. Receiver operating characteristic (ROC) curve plot of the
surgery.16,17 Prior revascularization might be a marker of the
preoperative prediction model and its derived risk score model.
severity of vascular disease and also indicate tissue ischemia or
confer higher risk if the index surgery was a revision of a prior
procedure, as revision surgeries have previously been the preoperative and perioperative period, such as chlor-
described to be higher risk for SSI after orthopedic proce- hexidine bathing, skin preparation, antibiotic prophylaxis,
dures.18,19 Whereas diabetes had a marginally significant as- and glucose control. Although it can be argued that all
sociation with SSI in the univariate analysis, it was not an SSI prevention measures should be universally adopted,
independent predictor in our multivariable analysis, in which real-world infection prevention practice is resource-
other risk factors had stronger associations. intensive and difficult to implement.20,21 Therefore, tar-
A clinical tool that quantifies SSI risk in this manner geted implementation of certain measures in high-risk
can potentially provide an impetus for strict adherence to patients could have potential to improve compliance with
evidence-based recommendations for SSI prevention in best practice, particularly in resource-limited settings.22

182 Leekha et al
Fig 3. Bubble plot of absolute risk of surgical site infection (SSI)
across the range of preoperative model risk scores.
In addition, rates of vascular SSI have remained high na-
tionally despite near-universal compliance with measures
outlined by the Surgical Care Improvement Project,23 sug-
gesting that additional patient-centered approaches might
be necessary. Knowledge of SSI risk could also help sur-
geons and patients more accurately weigh the risks and
benefits of options, such as use of prosthetic material and
nonoperative interventions. Providers can also be more
vigilant in monitoring high-risk patients for development
of SSI in the postoperative period. Last, the benefit of addi-
tional measures, such as supplemental oxygenation and
maintenance of normothermia, which has been found to
be helpful in decreasing SSI risk in certain surgeries,24
needs to be studied in this population, in which critical
limb ischemia and COPD were strong independent predic-
tors of SSI.
Other studies have developed prediction scores for SSI
after specific surgical procedures. In a study of patients in
the Society of Thoracic Surgeons National Cardiac
Database undergoing coronary artery bypass grafting,7
Fowler et al developed two different scoring systems for
major infection risk after surgery: one based on preopera-
tive factors alone, and the other using a combination of
preoperative and intraoperative factors. Notably, inclusion
of intraoperative factors added minimal predictive ability
to their model using preoperative factors alone. In a study
conducted in French hospitals among patients undergoing
abdominal noncolorectal surgery,2 Pessaux et al used pre-
operative variables to develop a scoring system to stratify
patients into those who would or would not benefit from
antimicrobial prophylaxis. In another multicenter study
using data from the National Surgical Quality Improve-
ment Program during a 3-year period,4 Neumayer et al
developed a scoring system that could be used by surgeons
to assess patient risk of SSI after general and vascular
surgery, using preoperative clinical and laboratory data.
This tool was found to be a better predictor of SSI
risk compared with the existing National Nosocomial
Infections Surveillance risk index, but the study lacked
JOURNAL OF VASCULAR SURGERY
July 2016
data on organ/space infection and surgeries with implant
placement. Similarly, van Walraven and Musselman
recently analyzed National Surgical Quality Improvement
Program data to develop a risk score to predict the risk
of SSI across a broad number of surgeries.25 Although their
score had a very good C statistic of 0.8, it uses both preop-
erative and intraoperative factors. Moreover, whereas the
overall discrimination of this score across various types of
surgical procedures is good, it is unclear how the score
would perform within specific procedures including
vascular surgery.
A second utility of risk stratification models is for
interhospital comparisons of SSI rates. As public reporting
of health care-associated infections is increasing, a number
of SSIs are being included in Centers for Medicare and
Medicaid Services’ hospital report cards as markers of
health care quality.26 In the past decade, it became amply
clear that not all surgical procedures were similar, and the
existing NHSN risk index was inadequate for risk stratifica-
tion across the spectrum of surgical procedures reported to
the NHSN.18 Investigators from the Centers for Disease
Control and Prevention attempted to improve risk stratifi-
cation for public reporting by developing procedure-
specific models based on NHSN data, including vascular
surgery.19 These updated models developed for abdominal
aortic aneurysm repair and peripheral vascular bypass
included had C statistics of 0.63 and 0.60, respectively,
corresponding to a limited discriminatory ability. Although
the models were based on a large data set from facilities
across the United States, the primary limitation was lack
of information on several important patient-specific charac-
teristics, such as those used in the present study. Our study
illustrates the importance of collecting more patient-
specific information for improving risk stratification.
Further, whereas the NHSN models may have utility in
interfacility comparison of publicly reported data, they
cannot be used for preoperative risk prediction.
Finally, studies among patients with chronic diseases
have shown that patient education and empowerment in-
terventions can improve management of several medical
conditions.27-29 However, patient engagement in infection
prevention has been studied to a limited extent, primarily
in reminding health care personnel to practice hand hy-
giene.30-32 A recent study suggests that patients would
like to be more informed about health care-associated in-
fections and appear willing to assist with the prevention
of infections in the hospital setting.33 A tool that predicts
patient-specific risk for infection after surgery could be
used to engage patients in infection risk modification
both before and after surgery.
Strengths of our study include utilization of prospectively
collected infection data in a large cohort and comprehensive
evaluation of a large number of preoperative variables as
predictive factors for SSI. Major limitations of this study
include the following: data are from a single center, we
were unable to blind investigators to case or control status
as that became apparent during chart review of historical
data, and missing data elements could have limited the
JOURNAL OF VASCULAR SURGERY
Volume 64, Number 1
assessment of specific candidate variables as predictive factors.
We could have missed cases of infections in patients who did
not return to our facility for management of infection; how-
ever, this is likely to be very small for cases requiring
hospitalization as those are generally documented by notifica-
tion of either the surgeon or the infection prevention and
control department. Our model is useful only for prediction
of more severe infections requiring hospital stay or
reoperation as we excluded those with superficial SSIs treated
with only oral antibiotics in the outpatient setting. Our data
are from an older cohort, when the prevalence of endovascu-
lar procedures was lower than in the present day; however, SSI
rates in vascular surgery remain high, and our study focused
on intrinsic patient-specific risk factors. Because we matched
on type of procedure, we were unable to assess the specific
type of procedure as a risk factor. Although it is likely that
different types of vascular surgeries have distinct risk factors,
because of relatively small numbers in each procedure
category, this study was not powered to develop specific
models for different procedures, such as abdominal aortic
aneurysm repair and peripheral vascular bypass. Future studies
should focus on prospective validation of this risk score in
external cohorts of different types of vascular procedures.
CONCLUSIONS
We have developed a simple SSI risk prediction score
for use before elective vascular surgery. This score requires
external validation in prospective cohorts before clinical
application but has potential to help patients and physicians
individualize and potentially modify SSI risk.
AUTHOR CONTRIBUTIONS
Conception and design: SL, BL, RT, PS, AD, RO
Analysis and interpretation: SL, BL, RT, PS, AD, RO
Data collection: SL, RT, PS, RO
Writing the article: SL, BL, RT, PS, AD, RO
Critical revision of the article: SL, BL, RO
Final approval of the article: SL, BL, RT, PS, AD, RO
Statistical analysis: BL
Obtained funding: SL, RO
Overall responsibility: SL
REFERENCES
1. Velasco E, Thuler LC, Martins CA, Dias LM, Goncalves VM. Risk
index for prediction of surgical site infection after oncology operations.
Am J Infect Control 1998;26:217-23.
2. Pessaux P, Atallah D, Lermite E, Msika S, Hay JM, Flamant Y, et al.
Risk factors for prediction of surgical site infections in “clean surgery”.
Am J Infect Control 2005;33:292-8.
3. Pessaux P, Lermite E, Blezel E, Msika S, Hay JM, Flamant Y, et al.
Predictive risk score for infection after inguinal hernia repair. Am J Surg
2006;192:165-71.
4. Neumayer L, Hosokawa P, Itani K, El-Tamer M, Henderson WG,
Khuri SF. Multivariable predictors of postoperative surgical site infec-
tion after general and vascular surgery: results from the Patient Safety in
Surgery Study. J Am Coll Surg 2007;204:1178-87.
5. Campos-Lobato LF, Wells B, Wick E, Pronty K, Kiran R, Remzi F,
et al. Predicting organ space surgical site infection with a nomogram.
J Gastrointest Surg 2009;13:1986-92.
Leekha et al 183
6. Chen LF, Anderson DJ, Kaye KS, Sexton DJ. Validating a 3-point
prediction rule for surgical site infection after coronary artery bypass
surgery. Infect Control Hosp Epidemiol 2010;31:64-8.
7. Fowler VG Jr, O’Brien SM, Muhlbaier LH, Corey GR, Ferguson TB,
Peterson ED. Clinical predictors of major infections after cardiac sur-
gery. Circulation 2005;112:I358-65.
8. Nashef SA, Roques F, Hammill BG, Peterson ED, Michel P,
Grover FL, et al. Validation of European System for Cardiac Operative
Risk Evaluation (EuroSCORE) in North American cardiac surgery. Eur
J Cardiothorac Surg 2002;22:101-5.
9. Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S,
Salamon R. European system for cardiac operative risk evaluation
(EuroSCORE). Eur J Cardiothorac Surg 1999;16:9-13.
10. Geubbels EL, Grobbee DE, Vandenbroucke-Grauls CM, Wille JC, de
Boer AS. Improved risk adjustment for comparison of surgical site
infection rates. Infect Control Hosp Epidemiol 2006;27:1330-9.
11. Friedman ND, Bull AL, Russo PL, Leder K, Reid C, Billah B, et al. An
alternative scoring system to predict risk for surgical site infection
complicating coronary artery bypass graft surgery. Infect Control Hosp
Epidemiol 2007;28:1162-8.
12. Berbari EF, Osmon DR, Lahr B, Eckel-Passow JE, Tsaras G,
Hanssen AD, et al. The Mayo prosthetic joint infection risk score:
implication for surgical site infection reporting and risk stratification.
Infect Control Hosp Epidemiol 2012;33:774-81.
13. National Healthcare Safety Network. Surveillance for surgical site
infection (SSI) events. Centers for Disease Control and Prevention
website. Available at: http://www.cdc.gov/nhsn/acute-care-hospital/
ssi/. Accessed February 12, 2015.
14. Ming DY, Chen LF, Miller BA, Sexton DJ, Anderson DJ. The impact
of depth of infection and postdischarge surveillance on rate of surgical-
site infections in a network of community hospitals. Infect Control
Hosp Epidemiol 2012;33:276-82.
15. National Quality Forum. National voluntary consensus standards for the
reporting of healthcare-associated infection data. Available at: http://
www.qualityforum.org/Publications/2008/03/National_Voluntary_
Consensus_Standards_for_the_Reporting_of_Healthcare-Associated_
Infection_Data.aspx. Accessed January 2, 2016.
16. Greenblatt DY, Rajamanickam V, Mell MW. Predictors of surgical site
infection after open lower extremity revascularization. J Vasc Surg
2011;54:433-9.
17. Kalish JA, Farber A, Homa K, Trinidad M, Beck A, Davies MG, et al.
Factors associated with surgical site infection after lower extremity
bypass in the Society for Vascular Surgery (SVS) Vascular Quality
Initiative (VQI). J Vasc Surg 2014;60:1238-46.
18. Leekha S, Sampathkumar P, Berry DJ, Thompson RL. Should national
standards for reporting surgical site infections distinguish between
primary and revision orthopedic surgeries? Infect Control Hosp Epi-
demiol 2010;31:503-8.
19. Mu Y, Edwards JR, Horan TC, Berrios-Torres SI, Fridkin SK.
Improving risk-adjusted measures of surgical site infection for the
National Healthcare Safety Network. Infect Control Hosp Epidemiol
2011;32:970-86.
20. Ranji SR, Shetty K, Posley KA, Lewis R, Sundaram V, Galvin CM, et al.
Closing the quality gap: a critical analysis of quality improvement
strategies (vol. 6: Prevention of healthcare-associated infections).
Rockville, Md: Agency for Healthcare Research and Quality (US);
2007. Report No. 04(07)-0051-6. AHRQ Technical Reviews.
21. Saint S, Howell JD, Krein SL. Implementation science: how to jump-start
infection prevention. Infect Control Hosp Epidemiol 2010;31:S14-7.
22. Harrod M, Manojlovich M, Kowalski CP, Saint S, Krein SL. Unique fac-
tors rural Veterans’ Affairs hospitals face when implementing health care-
associated infection prevention initiatives. J Rural Health 2014;30:17-26.
23. Dua A, Desai SS, Seabrook GR, Brown KR, Lewis BD, Rossi PJ, et al.
The effect of Surgical Care Improvement Project measures on national
trends on surgical site infections in open vascular procedures. J Vasc
Surg 2014;60:1635-9.
24. Anderson DJ, Podgorny K, Berríos-Torres SI, Bratzler DW,
Dellinger EP, Greene L, et al. Strategies to prevent surgical site in-
fections in acute care hospitals: 2014 update. Infect Control Hosp
Epidemiol 2014;35:S66-88.

184 Leekha et al
25. van Walraven C, Musselman R. The Surgical Site Infection Risk Score
(SSIRS): a model to predict the risk of surgical site infections. PloS One
2013;8:e67167.
26. Medicare.gov. Hospital Compare website. Healthcare-associated in-
fections. Available at: http://www.medicare.gov/hospitalcompare/
Data/Healthcare-Associated-Infections.html. Accessed May 14, 2015.
27. Effing T, Monninkhof EM, van der Valk PD, van der Palen J, van
Herwaarden CL, Partridge MR, et al. Self-management education for
patients with chronic obstructive pulmonary disease. Cochrane Data-
base Syst Rev 2007;4:CD002990.
28. Goeppinger J, Armstrong B, Schwartz T, Ensley D, Brady TJ.
Self-management education for persons with arthritis: managing
comorbidity and eliminating health disparities. Arthritis Rheum
2007;57:1081-8.
29. Grover SA, Lowensteyn I, Joseph L, Kaouache M, Marchand S,
Coupal L, et al. Patient knowledge of coronary risk profile improves the
effectiveness of dyslipidemia therapy: the CHECK-UP study: a ran-
domized controlled trial. Arch Intern Med 2007;167:2296-303.
JOURNAL OF VASCULAR SURGERY
July 2016
30. Davis RE, Sevdalis N, Pinto A, Darzi A, Vincent CA. Patients’ attitudes
towards patient involvement in safety interventions: results of two
exploratory studies. Health Expect 2013;16:e164-76.
31. McGuckin M, Taylor A, Martin V, Porten L, Salcido R. Evaluation of a
patient education model for increasing hand hygiene compliance in an
inpatient rehabilitation unit. Am J Infect Control 2004;32:235-8.
32. Stone S, Slade R, Fuller C, Charlett A, Cookson B, Teare L, et al. Early
communication: does a national campaign to improve hand hygiene
in the NHS work? Initial English and Welsh experience from the
NOSEC study (National Observational Study to Evaluate the Clean
YourHandsCampaign). J Hosp Infect 2007;66:293-6.
33. Seale H, Chughtai AA, Kaur R, Crowe P, Phillipson L, Novytska Y,
et al. Ask, speak up, and be proactive: empowering patient infection
control to prevent health care-acquired infections. Am J Infect Control
2015;43:447-53.
Submitted Sep 28, 2015; accepted Jan 11, 2016.