Hospital-Associated Costs Due to Surgical Site Infection After Breast ...

ORIGINAL ARTICLE 

Hospital-Associated Costs Due to Surgical Site 

Infection After Breast Surgery 

Margaret A. Olsen, PhD, MPH; Sorawuth Chu-Ongsakul, MD, MSc, MHA; Keith E. Brandt, MD; 

Jill R. Dietz, MD; Jennie Mayfield, BSN, MPH, CIC; Victoria J. Fraser, MD 

Objective: To determine the attributable costs associated 

with surgical site infection (SSI) following breast 

surgery. 

Design and Setting: Cost analysis of a retrospective 

cohort in a tertiary care university hospital. 

Patients: All persons who underwent breast surgery other 

than breast-conserving surgery from July 1, 1999, through 

June 30, 2002. 

Main Outcome Measures: Surgical site infection and 

hospital costs. Costs included all those incurred in the 

original surgical admission and any readmission(s) within 

1 year of surgery, inflation adjusted to US dollars in 2004. 

Results: Surgical site infection was identified in 50 

women during the original surgical admission or at readmission 

to the hospital within 1 year of surgery 

(N=949). The incidence of SSI was 12.4% following mastectomy 

with immediate implant reconstruction, 6.2% following 

mastectomy with immediate reconstruction using 

a transverse rectus abdominis myocutaneous flap, 4.4% 

following mastectomy only, and 1.1% following breast 

reduction surgery. Of the SSI cases, 96.0% were identified 

at readmission to the hospital. Patients with SSI had 

crude median costs of $16 882 compared with $6123 for 

uninfected patients. After adjusting for the type of surgical 

procedure(s), breast cancer stage, and other variables 

associated with significantly increased costs using 

feasible generalized least squares, the attributable cost of 

SSI after breast surgery was $4091 (95% confidence interval, 

$2839-$5533). 

Conclusions: Surgical site infection after breast cancer 

surgical procedures was more common than expected for 

clean surgery and more common than SSI after non– 

cancer-related breast surgical procedures. Knowledge of 

the attributable costs of SSI in this patient population can 

be used to justify infection control interventions to reduce 

the risk of infection. 

Arch Surg. 2008;143(1):53-60 

Author Affiliations: Division of 

Infectious Diseases, Department 

of Medicine (Drs Olsen, 

Chu-Ongsakul, and Fraser), 

and Divisions of Plastic and 

Reconstructive Surgery 

(Dr Brandt) and General 

Surgery (Dr Dietz), Department 

of Surgery, Washington 

University School of Medicine, 

and Infection Control 

Department, Barnes-Jewish 

Hospital (Dr Mayfield), 

St Louis, Missouri. 

Dr Chu-Ongsakul is now with 

the Department of Surgery, 

Siriraj Hospital Faculty of 

Medicine, Mahidol University, 

Bangkok, Thailand. Dr Dietz is 

now with the Department of 

General Surgery, Cleveland 

Clinic, Cleveland, Ohio. 

THE OVERALL SURGICAL SITE 

infection (SSI) rate following 

mastectomy reported to 

the Centers for Disease 

Control and Prevention National 

Nosocomial Infections Surveillance 

System from 1992 through 2004 was 

1.98%. 1 In contrast, SSI rates published in 

individual reports in the literature range 

See Invited Critique 

at end of article 

from 1% to 28%. Surgical site infection 

rates have been reported to be lowest following 

breast-conserving surgery, ranging 

from 1.5% to 10.8%. 2-8 The incidence 

of SSI following mastectomy ranges from 

2.8% to 25% in the surgical and infection 

control literature, 4-7,9-18 although only 2 

studies 4,5 report an incidence of less than 

5%. The SSI rate following breast cancer 

reconstructive surgery also seems to be 

relatively high (range, 6.3%-28%), based 

on a few reports in the literature. 10,19-22 

Thus, the SSI rates following breast surgery 

seem to be much greater than the nationally 

reported incidence of 2.0% and 

much higher than what is expected for 

clean surgical procedures. 

Given the state of fiscal constraints 

within the US health care system, it is important 

to calculate the cost-effectiveness 

of infection control interventions to 

justify their use from an economic perspective. 

Cost-effectiveness analyses require 

accurate estimates for the attributable 

costs of hospital-acquired infections, 

which are lacking for SSI. To our knowledge, 

there is only one report in the literature 

describing the attributable costs of 

SSI following breast surgery. Reilly et al 23 

estimated the costs associated with SSI af- 

(REPRINTED) ARCH SURG/ VOL 143 (NO. 1), JAN 2008 WWW.ARCHSURG.COM 

53 

Downloaded from www.archsurg.com at Washington University - St Louis, on January 21, 2008 

©2008 American Medical Association. All rights reserved.

ter a variety of different surgical procedures, including 

breast surgery. The costs included in this study were those 

associated with additional hospital length of stay, outpatient 

and home health visits, and outpatient antibiotics. 

Using the total costs attributable to infection presented 

in their publication, the mean attributable costs 

of SSI following breast surgery equaled £359 (approximately 

$574 in the United States). 23 It is not clear what 

types of breast surgery were included in the analysis (mastectomy, 

breast-conserving surgery, or cosmetic surgical 

procedures) or what proportion of the observed SSI 

were diagnosed and treated solely in the outpatient setting. 

Infections managed exclusively in an outpatient setting 

will presumably be associated with lower costs than 

infections requiring rehospitalization. In addition, British 

health care delivery and use are different from those 

in the United States, so the costs attributable to SSI after 

breast surgery determined in this study may not be comparable 

to the costs of SSI in the United States. 

Despite the publication of a number of studies describing 

costs associated with SSI, the magnitude of economic 

costs attributable to these infections is still unclear. 

Costs associated with SSI will likely vary depending 

on the type of surgical procedure. Surgical site infections 

following cardiothoracic or neurosurgical procedures, 

with the potential for severe morbidity and mortality, 

would presumably have the largest attributable 

costs, while infection following surgery not involving major 

organ spaces would be expected to have lower costs. 

This is consistent with the few adjusted estimates of SSI 

costs that have been reported. 23-30 We studied the hospitalassociated 

costs of SSI following mastectomy and breast 

reconstructive surgery to determine more precisely the 

attributable costs of SSI following these procedures. 

METHODS 

STUDY POPULATION 

Procedures eligible for inclusion in this study included all mastectomy 

and breast reconstructive surgical procedures performed 

at Barnes-Jewish Hospital (BJH) from July 1, 1999, through 

June 30, 2002. Barnes-Jewish Hospital is a 1251-bed tertiary care 

hospital affiliated with Washington University School of Medicine. 

International Classification of Diseases, Ninth Revision (ICD-9) 

procedure codes were used to identify eligible surgical admissions 

and outpatient surgical procedures, including breast reduction, 

augmentation, mastectomy, breast reconstruction with transverse 

rectus abdominis myocutaneous (TRAM) or other muscle 

flap, and insertion of a breast implant (ICD-9 procedure codes 

85.31-85.48, 85.7, 85.85, and 85.95). Admissions for breastconserving 

surgery (excisional biopsy, lumpectomy, or partial mastectomy) 

and mastopexy only were also excluded, because inpatient 

readmission for therapy of SSI after these procedures is rare. 

Using these criteria, 949 admissions with breast surgery were included 

in the original cohort. Approval for this study was obtained 

from the Washington University School of Medicine Human 

Studies Committee. 

IDENTIFICATION OF SSI CASE PATIENTS 

Surgical site infections following breast surgery were identified 

by electronic surveillance using data from the BJH Medical 

Informatics database. Indicators suggestive of SSI included 

ICD-9 diagnosis codes consistent with wound infection (codes 

682.2, 682.3, 998.5, 998.51, 998.59, and 996.69) or breast surgery 

complication (codes 611.0, 996.79, 998.3, and 998.83) 

and/or positive microbiology wound culture results during the 

original surgical admission or any readmission (inpatient or outpatient 

surgical) within 1 year of surgery. All potential wound 

infections identified by electronic surveillance were verified by 

review of the medical records. Centers for Disease Control and 

Prevention–National Nosocomial Infections Surveillance System 

criteria were used to define SSI. Surgical site infection required 

at least one of the following: purulent drainage from the 

incision, organisms isolated from an aseptically obtained culture, 

peri-incisional erythema or warmth and the incision was 

opened by a physician, abscess found on radiologic or histopathologic 

examination, or a diagnosis of SSI by the surgeon. 

Only infections diagnosed during the original surgical admission 

or at readmission to the hospital or to outpatient surgery 

were included; infections diagnosed and managed solely in the 

outpatient setting were excluded, because by definition outpatient 

infections would not have associated hospital costs. 

VARIABLE CREATION AND DEFINITIONS 

Data regarding variables potentially associated with increased 

costs were abstracted for the cohort from the BJH Medical Informatics 

database. The data available for the entire cohort included 

ICD-9 discharge diagnosis and procedure codes, pharmacy 

and microbiology data, and demographic information. 

Preexisting comorbidities, carcinoma in situ, malignancy, and 

history of malignancy were defined using the ICD-9 diagnosis 

codes assigned during the original surgical admission and the 

coding criteria of Deyo et al. 31 Comorbidities selected for analysis 

included those generally associated with increased risk of 

SSI (eg, diabetes mellitus) and comorbidities that may be more 

frequent in the population of patients with SSI and would most 

likely be associated with increased costs (eg, serious heart disease). 

In addition, frequencies were determined for all ICD-9 

diagnosis codes assigned to the surgical cohort; all comorbidities 

present in more than 1% of the cohort were included in 

the statistical analyses. The comorbidities analyzed included 

diabetes mellitus, serious heart disease (myocardial infarction 

or congestive heart failure), renal failure, chronic obstructive 

pulmonary disease, and hypertension. Diagnosis of metastatic 

disease was defined using ICD-9 diagnosis codes assigned during 

the original surgical admission and any readmission to the 

hospital within 1 year of surgery. Central venous catheterization 

was defined using ICD-9 procedure codes (code 38.93 or 

86.07) for the original surgical admission and any hospital readmission 

within 1 year of surgery. Surgical procedures performed 

during the original surgical admission were defined using 

ICD-9 procedure codes. All possible logic checks were performed 

for variables created from ICD-9 diagnosis and procedure 

codes, and discrepant results were verified using the electronic 

medical record. 

The number of ICD-9 codes assigned during the surgical admission 

was used as an indicator of severity of illness rather 

than the Charlson comorbidity score (adapted for use with administrative 

data), to determine the costs associated with specific 

comorbidities in this population. The Charlson comorbidity 

score is a summary score in which weights are assigned 

to various comorbidities associated with mortality; because our 

intent was to develop a precise determination of attributable 

costs, we used the individual comorbidities rather than a summary 

score in the statistical models to calculate costs. 32 In addition, 

the Charlson comorbidity score includes diagnosis of 

any malignancy in the scoring scheme. Because we wanted to 


54 



specifically control for costs associated with tumor stage in the 

patients diagnosed as having in situ or invasive breast cancer 

in our statistical models, this was an additional rationale for 

not using an aggregated comorbidity score in our models to adjust 

for severity of illness. 

Pharmacy data available from the BJH Medical Informatics 

database were also used to define some underlying disease variables, 

including insulin use in the original surgical admission 

or within 1 year of surgery, insulin and oral hypoglycemic agents 

as markers of diabetes mellitus, and digoxin use as a marker of 

serious heart disease. These variables were compared with the 

comorbidity variables previously described, to verify the diagnostic 

variables and to assess the effect of variables created by 

combination of the diagnostic and pharmacy information. 

COST DATA 

Cost data for the cohort from the perspective of the hospital, 

including the hospital costs of the original surgical admission 

and hospital costs for all readmissions within 1 year of surgery, 

were obtained from the BJH cost accounting database 

(Trendstar; McKesson Corp, Alpharetta, Georgia). Eligible hospital 

readmissions included inpatient, same day, or ambulatory 

surgery and outpatient surgery admissions. All readmission 

hospital costs within 1 year of surgery were included in 

the analyses, to hopefully capture all costs involved in the initial 

diagnosis and therapy of infection. Because SSI in some patients 

may require multiple admissions to the hospital for antibiotic 

or surgical treatment, we included readmission costs 

for all members of the cohort for a 1-year follow-up to capture 

these costs. 

The cost accounting database uses a set-down allocation 

method to calculate costs, which included indirect variable, direct 

variable, and fixed costs. All patient charge codes received 

during hospitalizations were recorded, and the departmental 

cost for each charge code calculated was based on each 

department’s actual cost components multiplied by the charges 

for that code, divided by total departmental charges. Costs were 

summed across each department to provide total hospital costs 

for each admission and were inflation adjusted to 2004 US dollars 

using the medical care component of the Consumer Price 

Index (accessed at http://www.bls.gov). Only hospitalassociated 

costs were included in the analysis; physician costs 

were not included. 

STATISTICAL ANALYSES 

The 2 test was used to compare categorical variables, and the 

Mann-Whitney test was used for continuous variables. Multivariate 

analyses to determine costs attributable to SSI were performed 

by ordinary least-squares and feasible generalized leastsquares 

(FGLS) regression, using natural log-transformed costs 

as the dependent variable, because of the highly skewed distribution 

of total costs. These methods can be used to estimate 

the costs associated with variables of interest, taking into account 

variation in the occurrence of other factors significantly 

associated with expenditures. The coefficients obtained in the 

FGLS model represent the mean change in the variable of interest 

per unit increase in X, when the other variables in the 

model are held constant. In this study, the change in X represents 

the mean difference in the natural logarithm of costs between 

individuals with infection and those without infection, 

holding all other predictors of costs constant. 

All independent variables were checked for colinearity. The 

models were checked for functional form misspecification using 

the Ramsey regression specification error test and for heteroscedasticity 

using the Breusch-Pagan test. 33 The total mortality 

within 1 year of surgery in the cohort was 1.5% (13/878); 

because it was so low, no attempt was made to adjust the FGLS 

models for mortality. 

Data for admissions in which the standardized residuals for 

the final model were large (3 SDs from the mean of 0) were 

all checked for accuracy. Because the FGLS model used the natural 

logarithm of costs as the dependent variable, an intermediate 

regression was performed to predict costs given that the 

dependent variable in the model was ln(costs). 33 

Propensity score methods were also used to determine the 

association between SSI and costs. 34 A nonparsimonious logistic 

regression model was used to obtain predicted values for 

SSI. Patients with SSI were then matched 1:1 with control subjects 

based on the propensity score. Total costs were compared 

between the matched pairs by the Wilcoxon signed rank 

test, and the mean and median differences in total costs were 

calculated. Statistical analyses were performed using commercially 

available software programs (SPSS, version 12.0 [SPSS 

Inc, Chicago, Illinois] and Stata, version 9.2 [Stata Corp, College 

Station, Texas]). 

RESULTS 

SSI RATES IN THE COHORT 

A cohort of 949 admissions in which ICD-9 procedure 

codes for mastectomy or breast reconstruction were assigned 

was established for the 2-year period of the study. 

Most surgical procedures (90.6%) resulted in at least an 

overnight hospitalization, with only 4 patients who underwent 

mastectomy (0.7%) discharged on the day of surgery. 

A total of 7.6% (14/185) of reduction-only surgical 

procedures, 50.8% (33/65) of delayed implant 

placement in patients with cancer, and 62.5% (15/24) of 

cosmetic augmentations were performed in outpatient surgery. 

Surgical site infection was identified in 50 patients 

within 1 year of surgery in this cohort (SSI rate, 5.3%). 

Thirteen infections (4.4%) followed mastectomy only, and 

22 were breast implant associated, requiring surgical removal 

of the implant in 18. Surgical site infection in 45 

patients involved the breast incision(s), 7 involved the 

abdominal incision in patients who had undergone TRAM 

reconstruction, and 1 involved the back incision in a woman 

who had undergone latissimus dorsi reconstruction. Three 

patients had SSI following TRAM reconstruction involving 

the breast and abdominal incisions. Surgical site infection 

rates were lowest following non–cancer-related 

surgical procedures. Rates of SSI according to the type of 

surgery performed are shown in Table 1. 

For the outcome and cost analyses, any readmissions 

to the hospital within 1 year of surgery, including readmissions 

in which eligible breast surgical procedures were 

performed, were included in the readmission costs associated 

with the original surgical admission. A total of 

61 patients had 1 or more follow-up breast surgical procedures 

performed within 1 year of the original surgery 

that were included in readmission costs, reducing the 

number of breast surgical admissions in the outcome cohort 

to 888. Ten patients had surgical admissions separated 

by more than 1 year; these were considered separate 

events. Forty patients (4.5%) included in the final 

breast surgical admission cohort were male; 33 males underwent 

subcutaneous mastectomies or reduction mam- 


55 



Table 1. Breast SSI Rates According to Type of Surgical 

Procedure Performed 

Type of Surgery a 

No. of 

Breast SSIs/ 

Total No. 

of Patients 

No. of 

Donor Site 

SSIs/Total No. 

of Patients 

maplasty because of gynecomastia, while 7 underwent 

unilateral mastectomies for surgical removal of a breast 

malignancy. 

CHARACTERISTICS OF PATIENTS 

WITH SSI AND OUTCOMES 

SSI 

Rate 

per 

Person 

Mastectomy only 13/296 NA 4.4 

Mastectomy plus immediate 15/121 NA 12.4 

implant b,c 

Mastectomy, immediate implant, 2/14 0/14 14.3 

plus latissimus dorsi flap d 

Delayed implant e,f 5/65 0/3 e 7.7 

Implant for augmentation g 0/24 NA NA 

Mastectomy plus immediate 6/162 j 7/162 j 6.2 j 

TRAM flap b,h,i 

Delayed TRAM reconstruction e 1/28 0/28 3.6 

Mastectomy plus latissimus dorsi 1/17 1/17 11.8 

flap (no implant) h 

Delayed latissimus dorsi flap e,k 0/10 0/10 NA 

Reduction surgery only l 2/185 NA 1.1 

Subcutaneous mastectomy 

because of gynecomastia 

0/33 NA NA 

Abbreviations: NA, data not applicable; SSI, surgical site infection; 

TRAM, transverse rectus abdominis myocutaneous. 

a There were 949 surgical procedures performed. 

b One patient had a tissue expander placed after a failed attempt at TRAM 

flap reconstruction. 

c Three patients also underwent contralateral reduction. 

d One patient also underwent contralateral reduction. 

e One patient had a delayed TRAM flap and placement of a tissue expander, 

and 2 patients had a delayed latissimus dorsi flap plus implant. 

f Three patients had contralateral reductions, and 10 patients also had 

nipple reconstruction. 

g One patient underwent contralateral reduction. 

h Two patients had a TRAM and latissimus dorsi flaps. 

i Four patients also underwent contralateral reductions. 

j Three patients had donor site and breast SSI. 

k Three patients underwent contralateral reduction. 

l Seven patients underwent unilateral reduction and contralateral nipple 

reconstruction. 

Characteristics identified in the surgical cohort and their 

associations with subsequent SSI are shown in Table 2. 

All of the patients with SSI following breast surgery were 

female, and there was no association between SSI and race, 

age, and malignant or metastatic breast cancer. Patients 

with SSI were significantly more likely to have an underlying 

diagnosis of diabetes mellitus, to be obese (defined 

by a body mass index [calculated as weight in kilograms 

divided by height in meters squared] 30), to 

have undergone mastectomy or a surgical procedure involving 

placement of a breast implant, and to have a central 

venous catheter placed within 1 year of surgery (excluding 

catheters placed to administer antibiotics due to 

SSI). Patients with SSI had significantly more ICD-9 diagnosis 

codes assigned during the breast surgical admission. 

Surgical site infection was significantly less likely 

in patients who had undergone only reduction surgery, 

compared with patients who had other procedures 

performed. 

Information on antibiotic prophylaxis and drain use 

and management for individual patients was not available, 

because those data were not maintained electronically. 

During the study period, the standard of care at our 

institution was to give all patients undergoing the included 

breast surgical procedures1gofcefazolinasprophylaxis 

within 1 hour before incision. In a separate analysis 

of a case-control subset at our institution that 

overlapped the period of this study, 88% of patients were 

given cefazolin prophylactically, with the remainder receiving 

vancomycin, a combination of ampicillin and sulbactam, 

clindamycin, or other antibiotics. Only 2% of patients 

in the case-control study were not given prophylactic 

antibiotics (M.A.O.; M. Lefta, BS; J.R.D.; K.E.B.; R. Aft, 

MD; R. Matthews, MPH; J.M.; and V.J.F.; unpublished 

data, 2006). Drains were placed routinely in the remaining 

breast tissue or mastectomy bed in patients who underwent 

mastectomy, flap reconstruction, or reduction, 

but were not used routinely in patients undergoing cosmetic 

augmentation or delayed reconstruction with an 

implant. Patients undergoing simple mastectomy typically 

had a single drain placed in the mastectomy bed, 

and patients undergoing modified radical mastectomy had 

2 drains, 1 in the mastectomy bed and 1 in the axilla. Patients 

undergoing reconstruction with a TRAM flap had 

2 drains placed in the abdomen and 1 in the reconstructed 

breast, and patients undergoing latissimus dorsi 

flap reconstruction had 1 drain placed in the back and 

1 in the reconstructed breast. Patients were discharged 

from the hospital with the drain(s) in place, and the 

drain(s) were generally removed between 7 and 21 days 

after surgery when the drainage was less than 1 ounce 

per 24 hours. 

The median time from surgery until diagnosis of SSI 

was 25 days (range, 2-315 days; mean [SD], 46.6 [62.0] 

days). Of the 50 patients with SSI, 41 (82.0%) had their 

infections diagnosed within 60 days of surgery. Only 2 

patients were diagnosed as having an SSI during the original 

surgical admission; almost three-quarters of infected 

patients (72.0%) were treated during a single readmission 

to the hospital (Table 3). Thirty patients with 

SSI (60.0%) underwent subsequent surgery because of 

infection: 18 involved breast implant removal, while 12 

involved debridement or incision and drainage of the 

wound in the operating room. Surgical site infections in 

the remaining 20 patients were treated with intravenous 

antibiotics with or without bedside drainage or debridement 

(Table 3). 

COSTS ASSOCIATED WITH SSI 

Complete cost data were not available for 4 patients, including 

1 with an SSI (following mastectomy plus immediate 

placement of an implant), resulting in a population 

of 884 surgical admissions available for analysis. 

The comparisons of the total costs and hospital length 

of stay for the surgical admission plus all readmissions 

within 1 year of surgery for the cohort are shown in 

Table 4. Patients with SSI had significantly higher hospital 

costs associated with surgery and during the 1-year 


56 



Table 2. Selected Characteristics of the Breast Surgical Cohort Population and Association With SSI 

Characteristic 

Patients With SSI 

(n=50) a 

Uninfected Patients 

(n=838) a OR (95% CI) P Value 

Female sex 50 (100.0) 798 (95.2) NA .16 

Nonwhite race 18 (36.0) 234 (27.9) 1.5 (0.8-2.6) .22 

Obesity (BMI, 30) 27 (54.0) 282 (33.7) 2.3 (1.3-4.1) .003 

Diabetes mellitus 12 (24.0) 80 (9.5) 3.0 (1.5-6.0) .001 

Renal failure 2 (4.0) 7 (0.8) 4.9 (1.0-24.5) .09 

Chronic obstructive pulmonary disease 2 (4.0) 33 (3.9) 1.0 (0.2-4.4) .99 

Serious heart disease 5 (10.0) 40 (4.8) 2.2 (0.8-5.9) .10 

Carcinoma in situ 9 (18.0) 86 (10.3) 1.9 (0.9-4.1) .09 

Malignancy 31 (62.0) 460 (54.9) 1.3 (0.7-2.4) .33 

Malignancy plus lymph node metastases 14 (28.0) 169 (20.2) 1.5 (0.8-2.9) .18 

Metastatic breast cancer 0 24 (2.9) .39 

Mastectomy 42 (84.0) 562 (67.1) 2.6 (1.2-5.6) .01 

Axillary dissection 29 (58.0) 412 (49.2) 1.4 (0.8-2.5) .22 

Free TRAM procedure 4 (8.0) 100 (11.9) 0.6 (0.2-1.8) .40 

Nonmicrovascular TRAM procedure 7 (14.0) 77 (9.2) 1.6 (0.7-3.7) .31 

Latissimus dorsi flap 4 (8.0) 36 (4.3) 1.9 (0.6-5.7) .28 

Implant 22 (44.0) 165 (19.7) 3.2 (1.8-5.7) .001 

Bilateral surgery 17 (34.0) 306 (36.5) 0.9 (0.5-1.6) .72 

Reduction surgery only 2 (4.0) 164 (19.6) 0.2 (0-0.7) .006 

Central venous catheter 11 (22.0) 59 (7.0) 3.7 (1.8-7.6) .001 

Age, mean (range), y 52.5 (27-82) 51.3 (14-90) NA .41 

No. of ICD-9 codes assigned during 

surgical admission, mean (range) 

3 (1-8) 2 (1-9) NA .001 

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CI, confidence interval; ICD-9, International 

Classification of Diseases, Ninth Revision (ICD-9); NA, data not applicable; OR, odds ratio; SSI, surgical site infection; TRAM, transverse rectus abdominis 

myocutaneous. 

a Data are given as number (percentage) of each group unless otherwise indicated. 

period after surgery compared with uninfected patients, 

and they had a significantly longer total length of hospital 

stay (P.001 for both, Mann-Whitney test). The 

crude median costs attributable to SSI totaled approximately 

$10 750, but, as can be seen in Table 4, the range 

of costs for the infected and uninfected control patients 

was broad. The crude median length of stay attributable 

to SSI equaled 4.3 days, although the range of length of 

stay was also broad for SSI case patients and uninfected 

control patients (Table 4). 

Feasible generalized least squares was used to calculate 

the attributable costs of SSI in the year after breast 

surgery. The natural logarithm of the total costs was used 

as the dependent variable in the regression model, because 

of the highly skewed nature of total costs. The operative 

variables associated with significantly increased 

costs included the type of donor flap used to reconstruct 

breast tissue (free or nonmicrovascular TRAM or 

latissimus dorsi flap), placement of an implant immediately 

following mastectomy, and bilateral surgery 

(Table 5). The diagnosis of malignant disease was associated 

with significantly increased costs, and metastatic 

disease was associated with even greater costs (indicated 

by the progressively larger values for the 

coefficient). Correspondingly, placement of a central venous 

catheter, used as a proxy for chemotherapy, was associated 

with increased costs. The only comorbidities associated 

with significantly increased costs in this 

population were serious heart disease (acute or old myocardial 

infarction, congestive heart failure, or digoxin administration 

in the hospital in the year after surgery) and 

Table 3. Outcomes of SSI in 50 Patients After 

Breast Surgery 

Characteristic 

No. (%) 

of Patients 

No. of readmissions required for therapy of SSI 

0 (diagnosed and treated solely during the 

2 (4.0) 

original surgical admission) 

1 36 (72.0) 

2 11 (22.0) 

3 1 (2.0) 

Type of surgery required to treat infection 

Removal of breast implant a 18 (36.0) 

Incision and drainage or debridement 

12 (24.0) 

in operating room b 

Treated with intravenous antibiotics only 20 (40.0) 

Abbreviation: SSI, surgical site infection. 

a One patient had the implant removed at an outside hospital because of 

infection and was subsequently treated at our institution with intravenous 

antibiotics. 

b Two patients with donor site SSI had abdominal mesh removed during 

surgery. 

renal failure. A diagnosis of male gynecomastia and reduction 

surgery in females was associated with significantly 

decreased costs. Controlling for the other variables 

in the model, costs for nonwhite persons were 7.9% 

higher in this breast surgical cohort compared with costs 

incurred by whites (Table 5). Excluding noncancer surgical 

procedures, diagnoses of breast cancer with lymph 

node metastases and metastatic breast cancer within 1 


57 



Table 4. Crude Costs and Length of Hospital Stay Associated With SSI in the Cohort a 

Cost Data 

LOS Data 

Infection Status 

Mean (SD), $ Median (Range), $ Mean (SD), d Median (Range), d 

Patients with SSI (n=49) b 19 901 (14 214) 16 882 (5198-67 763) 8.4 (7.4) 6.3 (1.4-39.3) 

Uninfected control patients (n=835) 9973 (10 013) 6123 (1455-120 037) 3.5 (4.4) 2.0 (0.2-47.5) 

Abbreviations: LOS, length of stay; SSI, surgical site infection. 

a The median cost and median LOS were used to calculate the crude cost ($10 759) and LOS (4.3 days) associated with SSI because of the skewed distribution 

of both variables. 

b Cost data were missing for 1 patient with an SSI. 

Table 5. Results of the Feasible Generalized Least-Squares Model for Determining Attributable Costs of Surgical Site Infection After 

Breast Surgery a 

Variable 

Proportion 

of Patients 

Estimated 

Coefficient 

Surgical site infection 0.06 .45 .06 .001 

Free TRAM procedure 0.12 1.12 .04 .001 

Nonmicrovascular TRAM procedure 0.09 .82 .04 .001 

Latissimus dorsi flap 0.04 .50 .07 .001 

Implant 0.21 .14 .08 .06 

Mastectomy plus immediate implant 0.15 .49 .09 .001 

Renal failure 0.009 1.28 .18 .001 

Serious heart disease 0.05 .52 .09 .001 

Diabetes mellitus 0.10 .03 .05 .48 

Metastatic cancer b 0.03 .96 .11 .001 

Central venous catheter c 0.08 .34 .06 .001 

Malignancy 0.55 .12 .04 .001 

Bilateral surgery 0.36 .19 .04 .001 

Nonwhite race 0.28 .08 .03 .007 

Weight, mean, kg 76.90 .001 .0006 .01 

Reduction 0.24 −.17 .05 .002 

Gynecomastia 0.04 −.43 .08 .001 

Abbreviation: TRAM, transverse rectus abdominis myocutaneous. 

a Data were also adjusted for number of International Classification of Diseases, Ninth Revision, Clinical Modification codes assigned during surgical admission 

(using dummy variables), age, weight, mastectomy, and year of surgery. Adjusted R 2 =0.52 for the model after accounting for natural log transformation of costs. 

b Diagnosed during the original admission or within 1 year of surgery. 

c Inserted within 1 year of surgery. 

SE 

P 

Value 

year of surgery were more common in nonwhite compared 

with white persons (31.7% vs 25.9% for lymph node 

metastases [P=.13] and 6.5% vs 2.5% for metastatic breast 

cancer [P=.01]), suggesting that nonwhites present at a 

more advanced stage of cancer than do whites. 

The attributable cost of SSI in this cohort, adjusted 

for the variables previously described and others listed 

in Table 5, was $4091 (95% confidence interval, $2839- 

$5533) (in 2004 US dollars). The final model presented 

in Table 5 had an adjusted coefficient of determination 

(R 2 ) of 0.52, indicating that approximately 52% of the 

variation in actual costs was explained by the model. 

In addition to the FGLS model, total costs were also compared 

between SSI cases and controls matched based on 

the propensity scores. A nonparsimonious logistic regression 

model was developed with SSI as the dependent variable. 

Uninfected control patients were matched to SSI case 

patients based on the propensity score, with 2 SSI case patients 

unable to be matched. The difference in total costs 

between the matched pairs was significant (P.001, Wilcoxon 

signed rank test). The mean difference in total costs 

between the matched SSI and control pairs was $5700, with 

a median difference in costs of $3492. 

COMMENT 

Knowledge of the outcomes of breast surgery, including 

SSI and their associated costs, is necessary to understand 

the impact of these infections in a population of surgical 

patients. Our study, using a retrospective cohort of breast 

surgical patients treated at a university-affiliated tertiary care 

hospital, shows that SSI after breast cancer surgery is more 

common than would be expected for clean surgery and more 

common than SSI rates after similarly extensive breast surgical 

procedures in patients without cancer. Given the relatively 

high incidence of SSI after these procedures, it is important 

to understand their associated costs, to develop 

prevention strategies and make informed decisions concerning 

the potential cost-effectiveness of interventions designed 

to reduce the risk of SSI. For this reason, it is important 

to quantify the costs attributable to infection as 


58 



precisely as possible, to determine the true magnitude of 

hospital-associated costs and, ultimately, societal costs because 

of SSI. 

In this study, we used FGLS to calculate attributable hospital 

costs of SSI after breast surgery, controlling for differences 

in surgical procedures, comorbidities, stage of breast 

cancer, and other factors associated with increased costs 

in this surgical population. We chose this method because 

it allows for the analysis of data from an entire cohort 

of surgical patients, in contrast to matched-pair studies 

in which case patients with SSI are matched to control 

patients without infection after surgery, resulting in analysis 

of only a selected subset of the control population. The 

inability to match some infected patients with unusual combinations 

of comorbidities or severe underlying illness with 

comparable control patients is an inherent disadvantage of 

the matched-pair method. The ordinary least-squares 

method using log-transformed costs as the dependent variable 

has recently been shown to accurately predict costs 

following coronary artery bypass graft surgery in a validation 

sample. 35 

Most published reports 36,37 addressing the costs of SSI 

have used a matched case-control study design, although 

this study design is known to result in potentially 

inaccurate cost estimates, because of selection bias. 

The estimated costs of SSI determined in studies in which 

some form of adjustment for patient acuity, and other 

factors associated with increased risk of SSI, were performed, 

either by matching or other statistical methods, 

range from approximately $3400 to $17 700. The adjusted 

costs of SSI have been lower following general or 

mixed surgical procedures ($3382-$5038) 26,30 and highest 

following orthopedic or cardiothoracic surgery 

($17 708 for orthopedic SSI and $12 419-$14 211 following 

coronary artery bypass surgery). 27-29 These cost 

estimates are not necessarily comparable, even for the 

same type of surgery, because of differing statistical methods 

to determine attributable costs, different duration of 

follow-up after surgery, and the inclusion of different 

sources of cost data in the various studies. 

In the present study, we calculated crude median costs 

attributable to SSI after breast surgical procedures equal 

to $10 759. After accounting for the variety of operative 

methods, diagnoses, and comorbidities, the adjusted attributable 

costs of SSI equaled $4091. We found similar 

results using propensity scores and calculation of median 

differences in costs between matched SSI cases and 

controls with similar propensity to develop an infection. 

Our cost estimates included costs incurred during 

the original surgical admission and all hospital readmissions 

within 1 year of surgery for all patients in the cohort. 

We included all readmission costs during a 1-year 

follow-up for several reasons. First, the standard definition 

of deep SSI used by the Centers for Disease Control 

and Prevention–National Nosocomial Infections Surveillance 

System includes infections with onset within 1 year 

of surgery in patients with foreign bodies implanted during 

surgery. Of the patients in our cohort, 21.0% had a 

breast implant placed during surgery and, thus, we needed 

to extend the period of observation to include infections 

associated with implants. Second, therapy due to 

SSI required multiple readmissions to the hospital in some 

patients, so we needed to include all of these readmissions 

to capture all costs associated with therapy due to 

SSI. We included readmission costs for control patients 

to model the costs directly attributable to SSI. For example, 

patients with SSI may also have had complications 

of chemotherapy during a readmission hospitalization, 

so not all of the costs incurred during that 

readmission were directly attributable to SSI. By including 

all readmission costs for case patients and controls, 

we could use the ordinary least-squares method to model 

the costs attributable to each of the variables in our model. 

The only underlying comorbidities associated with increased 

costs in this patient population were serious heart 

disease and renal failure, most likely due to the occurrence 

of breast cancer in primarily elderly women. Nonwhite 

race was also associated with significantly increased 

costs in our ordinary least-squares model, possibly 

because of later stage at diagnosis of breast cancer in African 

American women compared with white women. 38 This 

is supported by the finding that more nonwhite persons 

in our cohort were diagnosed as having metastatic breast 

cancer within 1 year of surgery than were white persons, 

consistent with more advanced disease at diagnosis 

in nonwhite persons. 

Our results are consistent with the hypothesis that SSI 

following surgery involving major organ spaces are associated 

with greater costs than SSI following less extensive 

surgical procedures. Our calculated attributable cost of 

$4091 is much less than the attributable costs of SSI reported 

after cardiothoracic surgery 24,27,28 and orthopedic surgery, 

29 but much higher, however, than the value reported 

by Reilly et al 23 (approximately $574) in breast 

surgical patients. The higher attributable cost associated 

with SSI that we calculated may be primarily because of 

the inclusion of different types of surgical procedures in 

our study compared with that of Reilly and colleagues. We 

excluded breast-conserving surgery from our cohort because 

of the rarity of inpatient readmission for therapy due 

to SSI in this population at our institution (M.A.O., J.R.D., 

and V.J.F., unpublished data, 2006). Reilly et al included 

costs incurred in the hospital and outpatient setting and, 

if their population included women undergoing breastconserving 

surgery, they may have been able to capture the 

outpatient costs associated with SSI following this procedure, 

resulting in the much lower value reported compared 

with our result. 

The attributable cost of SSI we calculated, $4091 per 

patient, includes only hospital-associated costs and, thus, 

certainly represents an underestimate of the true societal 

costs of serious SSI in this surgical population. Our 

study did not include physician costs or costs due to SSI 

associated with excess clinic visits, outpatient surgical 

procedures performed in the general or plastic surgery 

clinics, outpatient antibiotic use, or home health care and, 

thus, the total attributable costs we calculated represent, 

at best, the minimum costs associated with serious 

SSI resulting in readmission to the hospital or outpatient 

surgery. 

Ultimately, the goal of studying the incidence and costs 

associated with SSI is to develop and implement interventions 

to reduce the occurrence of this adverse surgical 

event. Potential interventions to reduce the inci- 


59 



dence of SSI in this patient population include strategies 

to optimize the timing and dosage of prophylactic antibiotics 

administered before the surgical incision, glucose 

control in diabetic patients, promotion of meticulous 

hand hygiene, and strategies to promote timely 

removal of drains, among others. Interventions to reduce 

the incidence of SSI following breast cancer surgical 

procedures are essential to reduce not only morbidity 

in these patient populations but also costs to the 

individuals and to society. 

Accepted for Publication: September 4, 2006. 

Correspondence: Margaret A. Olsen, PhD, MPH, Division 

of Infectious Diseases, Washington University School 

of Medicine, 660 S Euclid Ave, Campus Box 8051, St 

Louis, MO 63110 (molsen@im.wustl.edu). 

Author Contributions: Study concept and design: Olsen, 

Brandt, Dietz, and Fraser. Acquisition of data: Olsen, Chu- 

Ongsakul, Mayfield, and Fraser. Analysis and interpretation 

of data: Olsen, Chu-Ongsakul, Dietz, and Fraser. Drafting 

of the manuscript: Olsen and Fraser. Critical revision of 

the manuscript for important intellectual content: Chu- 

Ongsakul, Brandt, Dietz, Mayfield, and Fraser. Statistical 

analysis: Olsen. Obtained funding: Fraser. Administrative, 

technical, and material support: Chu-Ongsakul, Mayfield, 

and Fraser. Study supervision: Brandt, Dietz, and Fraser. 

Financial Disclosure: None reported. 

Funding/Support: This study was supported by Epicenter 

Prevention Program Cooperative Agreement VR8/ 

CCU715087 from the Centers for Disease Control and 

Prevention (Dr Fraser). 

Additional Contributions: Christopher S. Hollenbeak, 

PhD, and Donald Nichols, PhD, provided insight and comments 

concerning the creation of the ordinary leastsquares 

and generalized least-squares cost models; and 

Barton Hamilton, PhD, provided suggestions concerning 

the use of propensity score methods. 

REFERENCES 

1. National Nosocomial Infections Surveillance System. National Nosocomial Infections 

Surveillance (NNIS) System Report, data summary from January 1992 through 

June 2004, issued October 2004. Am J Infect Control. 2004;32(8):470-485. 

2. Seltzer MH. Partial mastectomy and limited axillary dissection performed as same 

day surgical procedure in the treatment of breast cancer. Int Surg. 1995;80 

(1):79-81. 

3. Keidan RD, Hoffman JP, Weese JL, et al. Delayed breast abscesses after lumpectomy 

and radiation therapy. Am Surg. 1990;56(7):440-444. 

4. Tejler G, Aspegren K. Complications and hospital stay after surgery for breast 

cancer: a prospective study of 385 patients. Br J Surg. 1985;72(7):542-544. 

5. Platt R, Zucker JR, Zaleznik DF, et al. Prophylaxis against wound infection following 

herniorrhaphy or breast surgery. J Infect Dis. 1992;166(3):556-560. 

6. Rotstein C, Ferguson R, Cummings KM, Piedmonte MR, Lucey J, Banish A. 

Determinants of clean surgical wound infections for breast procedures at an oncology 

center. Infect Control Hosp Epidemiol. 1992;13(4):207-214. 

7. Tran CL, Langer S, Broderick-Villa G, DiFronzo LA. Does reoperation predispose 

to postoperative wound infection in women undergoing operation for breast cancer? 

Am Surg. 2003;69(10):852-856. 

8. Bertin ML, Crowe J, Gordon SM. Determinants of surgical site infection after breast 

surgery. Am J Infect Control. 1998;26(1):61-65. 

9. Chen J, Gutkin Z, Bawnik J. Postoperative infections in breast surgery. J Hosp 

Infect. 1991;17(1):61-65. 

10. Vinton AL, Traverso LW, Zehring RD. Immediate breast reconstruction following 

mastectomy is as safe as mastectomy alone. Arch Surg. 1990;125(10): 

1303-1308. 

11. Morimoto K, Kinoshita H. Once-daily use of ofloxacin for prophylaxis in breast 

cancer surgery. Chemotherapy. 1998;44(2):135-141. 

12. Broadwater JR, Edwards MJ, Kuglen C, Hortobagyi GN, Ames FC, Balch CM. 

Mastectomy following preoperative chemotherapy: strict operative criteria control 

operative morbidity. Ann Surg. 1991;213(2):126-129. 

13. Beatty JD, Robinson GV, Zaia JA, et al. A prospective analysis of nosocomial wound 

infection after mastectomy. Arch Surg. 1983;118(12):1421-1424. 

14. Lipshy KA, Neifeld JP, Boyle RM, et al. Complications of mastectomy and their 

relationship to biopsy technique. Ann Surg Oncol. 1996;3(3):290-294. 

15. Nieto A, Lozano M, Moro MT, Keller J, Carralafuente C. Determinants of wound 

infections after surgery for breast cancer. Zentralbl Gynakol. 2002;124(8-9): 

429-433. 

16. Hayes JA, Bryan RM. Wound healing following mastectomy. AustNZJSurg. 

1984;54(1):25-27. 

17. Badr el Din A, Coibion M, Guenier C, et al. Local postoperative morbidity following 

pre-operative irradiation in locally advanced breast cancer. Eur J Surg Oncol. 

1989;15(6):486-489. 

18. Say CC, Donegan W. A biostatistical evaluation of complications from mastectomy. 

Surg Gynecol Obstet. 1974;138(3):370-376. 

19. Alderman AK, Wilkins EG, Kim HM, Lowery JC. Complications in postmastectomy 

breast reconstruction: two-year results of the Michigan Breast Reconstruction 

Outcome Study. Plast Reconstr Surg. 2002;109(7):2265-2274. 

20. Vandeweyer E, Deraemaecker R, Nogaret JM, Hertens D. Immediate breast reconstruction 

with implants and adjuvant chemotherapy: a good option? Acta Chir 

Belg. 2003;103(1):98-101. 

21. Van Geel AN, Contant CM, Wai RT, Schmitz PI, Eggermont AM, Menke- 

Pluijmers MM. Mastectomy by inverted drip incision and immediate reconstruction: 

data from 510 cases. Ann Surg Oncol. 2003;10(4):389-395. 

22. Nahabedian MY, Tsangaris T, Momen B, Manson PN. Infectious complications 

following breast reconstruction with expanders and implants. Plast Reconstr Surg. 

2003;112(2):467-476. 

23. Reilly J, Twaddle S, McIntosh J, Kean L. An economic analysis of surgical wound 

infection. J Hosp Infect. 2001;49(4):245-249. 

24. Boyce JM, Potter-Bynoe G, Dziobek L. Hospital reimbursement patterns among 

patients with surgical wound infections following open heart surgery. Infect Control 

Hosp Epidemiol. 1990;11(2):89-93. 

25. Zoutman D, McDonald S, Vethanayagan D. Total and attributable costs of surgicalwound 

infections at a Canadian tertiary-care center. Infect Control Hosp Epidemiol. 

1998;19(4):254-259. 

26. Kirkland KB, Briggs JP, Trivette SL, Wilkinson WE, Sexton DJ. The impact of surgicalsite 

infections in the 1990s: attributable mortality, excess length of hospitalization, 

and extra costs. Infect Control Hosp Epidemiol. 1999;20(11):725-730. 

27. Jenney AWJ, Harrington GA, Russo PL, Spelman DW. Cost of surgical site infections 

following coronary artery bypass surgery. ANZ J Surg. 2001;71(11): 

662-664. 

28. Hollenbeak CS, Murphy D, Dunagan WC, Fraser VJ. Nonrandom selection and 

the attributable cost of surgical-site infections. Infect Control Hosp Epidemiol. 

2002;23(4):177-182. 

29. Whitehouse JD, Friedman D, Kirkland KS, Richardson WJ, Sexton DJ. The impact 

of surgical-site infections following orthopedic surgery at a community hospital 

and a university hospital: adverse quality of life, excess length of stay, and 

extra cost. Infect Control Hosp Epidemiol. 2002;23(4):183-189. 

30. Perencevich EN, Sands KE, Cosgrove SE, Guadagnoli E, Meara E, Platt R. Health 

and economic impact of surgical site infections diagnosed after hospital discharge. 

Emerg Infect Dis. 2003;9(2):196-203. 

31. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with 

ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613-619. 

32. Charlson ME, Pompei P, Ales K, MacKenzie CR. A new method of classifying prognostic 

comorbidity in longitudinal studies: development and validation. J Chronic 

Dis. 1987;40(5):373-383. 

33. Wooldridge JM. Introductory Econometrics: A Modern Approach. Mason, OH: 

Thomson Learning; 2003. 

34. Rubin DB. Estimating causal effects from large data sets using propensity scores. 

Ann Intern Med. 1997;127(8, pt 2):757-763. 

35. Austin PC, Ghali WA, Tu JV. A comparison of several regression models for analysing 

cost of CABG surgery. Stat Med. 2003;22(17):2799-2815. 

36. Smith HL. Matching with multiple controls to estimate treatment effects in observational 

studies. Sociol Methodol. 1997;27:325-353. 

37. Ming K, Rosenbaum PR. Substantial gains in bias reduction from matching with 

a variable number of controls. Biometrics. 2000;56(1):118-124. 

38. Ghafoor A, Jemal A, Ward E, Cokkinides V, Smith R, Thun M. Trends in breast 

cancer by race and ethnicity. CA Cancer J Clin. 2003;53(6):342-355. 


60

Hospital-Associated Costs Due to Surgical Site Infection After Breast ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?