1247

Risk Factors for Prosthetic Joint Infection: Case-Control Study

Elie F. Berbari, Arlen D. Hanssen, Mary C. Duffy,
James M. Steckelberg, Duane M. Ilstrup,
William S. Harmsen, and Douglas R. Osmon

From the Division of Infectious Diseases, the Department of

Orthopedics, and the Section of Biostatistics, Mayo Clinic and Mayo
Foundation, Rochester, Minnesota

We conducted a matched case-control study to determine risk factors for the development of
prosthetic joint infection. Cases were patients with prosthetic hip or knee joint infection. Controls
were patients who underwent total hip or knee arthroplasty and did not develop prosthetic joint
infection. A multiple logistic regression model indicated that risk factors for prosthetic joint infection
were the development of a surgical site infection not involving the prosthesis (odds ratio [OR], 35.9;
95% confidence interval [CI], 8.3 154.6), a National Nosocomial Infections Surveillance (NNIS)
System surgical patient risk index score of 1 (OR, 1.7; 95% CI, 1.2 2.3) or 2 (OR, 3.9; 95% CI, 2.0
7.5), the presence of a malignancy (OR, 3.1; 95% CI, 1.3 7.2), and a history of joint arthroplasty
(OR, 2.0; 95% CI, 1.4 3.0). Our findings suggest that a surgical site infection not involving the joint
prosthesis, an NNIS System surgical patient risk index score of 1 or 2, the presence of a malignancy,
and a history of a joint arthroplasty are associated with an increased risk of prosthetic joint infection.

control study to identify patient populations at increased risk

of prosthetic joint infection.

Methods
Patient Population

The Mayo Clinic (Rochester, MN) is a tertiary care academic

medical center in the upper Midwest of the United States;
340,000 patients per year are cared for at this center. The study
population consisted of all patients without a history of prosthetic
joint infection who had a THA or TKA implanted at the Mayo
Clinic between 1 January 1969 and 31 December 1991.
Cases of prosthetic joint infection at the Mayo Clinic were
identified by using the total joint arthroplasty registry [19],
the master diagnostic index [20], and clinical microbiology
laboratory data from 1969 through 1991. Follow-up data for
all patients were obtained by using a computerized database
compiled by the total joint arthroplasty registry associated with
the Department of Orthopedic Surgery at the Mayo Clinic. This
registry is part of an ongoing prospective cohort study that
began in 1969 to investigate the complications and outcomes
of prosthetic joint implantation.
All patients were followed up at regular intervals after arthroplasty by examination, letter, or telephone contact. Minimum
follow-up was twice in the first postsurgical year and then
every 5 years thereafter. When follow-up examination at the
Mayo Clinic was not possible, the patient completed a standardized data collection form, and roentgenograms of the joint
prosthesis were sent to the Mayo Clinic for review. By using
this system, 90% of this cohort has been followed up.

Received 23 February 1998; revised 9 July 1998.

Reprints or correspondence: Dr. Douglas R. Osmon, Division of Infectious
Diseases, Department of Internal Medicine, Mayo Clinic/Mayo Foundation,
200 First Street Southwest, Rochester, Minnesota 55905.

Cases and Controls

Clinical Infectious Diseases 1998;27:124754

q 1998 by the Infectious Diseases Society of America. All rights reserved.
10584838/98/27050022$03.00

Cases were patients within the study population who developed a prosthetic joint infection. Only the first episode of pros-

/ 9c5b$$no30

10-14-98 12:06:56

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

Although prosthetic joint infection is rare following the estimated 430,000 total hip arthroplasty (THA) and total knee
arthroplasty (TKA) procedures performed in the United States
each year [1 9], it remains one of the major complications
that may lead to prosthesis removal or loss of function and is
associated with a mortality rate of 2.7% 18% [10, 11]. Treatment often requires removal of the infected prosthesis and
prolonged intravenous antimicrobial therapy. The cost of each
episode is estimated to be $50,000 [12]. Prevention of prosthetic joint infection includes augmentation of the host response, optimizing the wound environment, and reduction of
bacterial deposition into the wound in the preoperative, intraoperative, and postoperative periods [13].
Investigators have attempted to define characteristics that
predispose patients to prosthetic joint infection by using formal
epidemiological methods. However, most studies have been
limited by methodological problems including emphasizing
case series rather than formal observational cohort or casecontrol studies, a lack of explicit case or risk factor definitions,
incomplete case ascertainment, selection biases, failure to account for differences in duration or completeness of followup, and insufficient statistical power [4, 8, 14 18].
Identification of patients at high risk for prosthetic joint
infection would allow for improved preoperative risk assessment, increase the index of suspicion of health care providers
for prosthetic joint infection in high-risk individuals, and identify patients for whom focused efforts at prevention are necessary. Therefore, we performed a retrospective, matched case-

1248

Berbari et al.

Table 1. Definitions of potential host, index arthroplasty, and postoperative risk factors for prosthetic joint infection.
Risk factor

Definition

Host risk factor

Rheumatoid arthritis
Diabetes mellitus
Malignancy

Joint malignancy

Steroid use

Chronic renal insufficiency*

Pyuria*
Bacteriuria*
Obesity

Potential Risk Factors

The medical records of case and controls were abstracted

for data on potential risk factors. Definitions of risk factors are
listed in table 1. The presence of any other risk factor was
documented by the clinicians caring for the patient at the time
of prosthesis implantation.

Malnutrition

Index arthroplasty risk factor

NNIS System surgical
patient risk index score
Blood transfusion

Data Collection and Statistical Analysis

The medical records of all cases and controls were abstracted

by one of the authors, by using a standardized data collection
tool for demographic information and data related to potential
host, index arthroplasty, and postoperative risk factors for prosthetic joint infection. A conditional logistic regression model
was used to estimate the odds ratio and 95% confidence interval
of a variable for predicting the risk of prosthetic joint infection
while retaining the matching of cases to controls [35]. The
variables of age and date of prosthesis implantation could not
be matched exactly and therefore were included as main effect
adjustments in all models.
Potential risk factors were first assessed in a univariate
model. The sign test for dichotomous variables and the signedrank test for continuous variables were used to analyze the
hypothesis that the odds ratio for each individual risk factor
was significantly different than 1. Candidate variables for the
multivariate model were variables with a P value of .15
for which there were 20 discordant pairs among cases and
controls.
A multivariate analysis utilizing a stepwise method of model
building was used along with the methods of forward and
backward selection to aid in model confirmation. All methods
yielded the same model. Possible interactions between main

/ 9c5b$$no30

10-14-98 12:06:56

P.M. surgery
Postoperative risk factor
Surgical site infection

Nosocomial infection

As defined by the American Rheumatism

Association [21]
As defined by the National Diabetes
Data Group [22]
All malignancies, excluding skin cancer
other than melanoma and tumors
involving the index joint, diagnosed
within 5 years before total joint
arthroplasty
All primary or metastatic malignancies
involving the site of the index
arthroplasty that are diagnosed within
5 years before total joint arthroplasty
Any form of systemic steroid therapy for
1 week in the year before total joint
arthroplasty
A creatinine clearance of 30 mL/min
[23]
10 WBCs per high-power field [24]
105 colonies of bacteria/mL
Weight 20% above ideal body weight
[25]
An albumin level of 34 g/L or an
absolute lymphocyte count of 1.5 1
109/L [26 28]
As defined by the CDC [29]
Any autologous or homologous blood
transfusion within 24 hours of
prosthesis implantation
Surgery that occurred after 12:00 P.M.
Surgical site infection that does not
involve the prosthesis as defined by
the CDC [30]
As defined by the CDC [31]

NOTE. CDC Centers for Disease Control and Prevention; NNIS

National Nosocomial Infections Surveillance.
* Within 30 days before total joint arthroplasty.

Uses information on the length of the surgical procedure, the American

Society of Anesthesiologists preoperative assessment score [32], and surgical
wound classification for each procedure [29, 30, 33, 34]. The T time or cutoff
point utilized in this study was a surgery time of 3 hours [33].

Occurred after prosthesis implantation and before the development of prosthetic joint infection or hospital discharge.

Includes urinary tract infection, pneumonia, bloodstream infection, gastrointestinal infection, and other infections.

effect variables and the matching variables were examined, as

were interactions among the main effect variables.
After identifying the best main effect multivariate model,
the bootstrap method was used to validate the model [36].
Variables selected in 70% of the 600 bootstrap samples were
retained. This method yielded a result identical to that of the
multivariate model.

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

thetic joint infection in a given patient was included in the

analysis. Prosthetic joint infection was diagnosed if two or
more cultures of joint aspirates or cultures of intraoperative
specimens yielded the same microorganism, purulence surrounding the prosthesis was observed at the time of debridement or removal of the prosthesis, acute inflammation consistent with infection was present during histopathologic
examination, or a sinus tract that communicated with the prosthesis was present.
By using a computer-generated random selection scheme,
controls were randomly selected from a list of patients within
the study population who did not develop prosthetic joint infection. Controls were matched 1:1 to cases by age, sex, prosthesis
location, and date of implantation. Controls were matched to
cases by date of implantation to control for changes in clinical
practices during the study period. In addition, the length of
follow-up for each control had to be equal to or greater than
the time from prosthesis implantation to diagnosis of prosthetic
joint infection for the matched case.

CID 1998;27 (November)

CID 1998;27 (November)

Risk Factors for Prosthetic Joint Infection

Table 2. Demographic characteristics of cases with prosthetic joint

infection and their matched controls who were seen at the Mayo
Clinic between 1969 and 1991.

Results
Patient Population

Case-Control Study

Joint location and gender were matched exactly between the

two groups. The largest difference in age and date of prosthesis
implantation between a case and a matched control was 6.6
and 2 years, respectively. Controls were followed up for a
median of 11.2 years (range, 0.05 27.2 years) without the
development of prosthetic joint infection.
Univariate Analysis

The host, index arthroplasty, and postoperative variables analyzed as potential risk factors for prosthetic joint infection are
listed along with the results of the univariate analysis in table

10-14-98 12:06:56

Demographic characteristic
Prosthesis location
Hip
Knee
Median age in y (range)
Sex
Male
Female
No. of surgeons
Implantation date
1969 1979
1980 1991
Underlying native joint
disease*
Degenerative joint disease
Rheumatoid arthritis
Fracture
Avascular necrosis
Native joint septic arthritis
Posttraumatic arthritis
Congenital hip dysplasia
Underlying tumor or
metastasis
Mixed conditions
Others
Type of arthroplasty
Hip
Cemented THA
Uncemented THA
Hybrid THA
Cemented bipolar
endoprosthesis
Uncemented bipolar
endoprosthesis
Uncemented Austin
Moore endoprosthesis
Endoprosthesis
Cup arthroplasty
Miscellaneous
Knee
Cemented TKA
Uncemented TKA
Unicompartmental knee
arthroplasty
Hinged TKA

Cases
(n 462)

Controls
(n 462)

263 (57)
199 (43)
61.5 (16 91)

263 (57)
199 (43)
61.8 (15 89)

231 (50)
231 (50)
36

231 (50)
231 (50)
35

269 (58)
193 (42)

269 (58)
193 (42)

207
90
50
29
24
16
15

259
68
36
28
10
17
16

(45)
(20)
(11)
(6)
(5)
(3)
(3)

10 (2)
4 (1)
17 (4)
263
209
15
3

(57)
(79)
(6)
(1)

(56)
(15)
(8)
(6)
(2)
(4)
(3)

2 (0.0)
5 (1)
21 (5)
263
222
14
3

(57)
(84)
(5)
(1)

15 (6)

8 (3)

2 (1)

2 (1)

4
7
3
5
199
155
8

(1)
(3)
(1)
(2)
(43)
(78)
(4)

1 (1)
35 (17)

1
4
2
7
199
173
8

(0.0)
(2)
(1)
(3)
(43)
(87)
(4)

11 (6)
7 (3)

NOTE. Unless stated otherwise, data are no. (%) of patients with indicated
characteristic. THA total hip arthroplasty; TKA total knee arthroplasty.
* Reason for arthroplasty (based on historical, physical, radiological, and
intraoperative findings).

4. Postoperative wound complications remained significant risk

factors even after adjusting for the differences in the length of
hospital stay and the time between prosthesis implantation and
hospital dismissal between cases and controls.
The median duration of rheumatoid arthritis before arthroplasty was 17.5 years (range, 1 50 years) for the cases and 15

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

Between 1 January 1969 and 31 December 1991, 468 prosthetic joint infections occurred in 466 (1.8%) of 26,505 patients
who received a THA or TKA at our institution. Six episodes
(1.3%) of prosthetic joint infection in six patients were excluded from the analysis because of incomplete medical records; therefore, 462 episodes of prosthetic joint infection in
460 patients were available for analysis.
Demographic characteristics of the cases and their matched
controls are shown in table 2, and the microbiological findings
for the cases are listed in table 3. The median duration between
joint arthroplasty and diagnosis of prosthetic joint infection
was 512 days (range, 3 7,131 days). Of the 462 episodes, 89
(19%) occurred within 0 90 days of prosthesis implantation;
185 (40%), within 90 days to 2 years of prosthesis implantation;
and 188 (41%), 2 years after prosthesis implantation. Staphylococcus aureus was the most common pathogen followed by
a polymicrobial etiology and coagulase-negative staphylococci.
The largest number of prostheses that any individual surgeon
implanted in cases and controls was 39 (8.4%) and 43 (9.3%),
respectively. Degenerative joint disease was the most common
reason for prosthesis implantation in both groups. Polymethylmethacrylate cement for fixing both components of the prosthesis was the most common prosthetic design for THA and TKA
in both cases and controls (table 2).
Antimicrobial prophylaxis administered before surgery was
utilized for 392 cases (84.8%) and 402 controls (87%). Cefazolin was the most commonly prescribed surgical prophylaxis
(204 cases [52%] and 219 controls [54%]) followed by methicillin (155 [40%] and 157 [39%], respectively), and others (33
[8%] and 26 [7%], respectively). The duration of antimicrobial
prophylaxis was 48 hours for all cases and controls.
Ultraviolet light was not utilized to sterilize the operating
room air for any of the patients in our study population. Horizontal laminar airflow rooms became available at our institution
in 1981. Antibiotic-impregnated bone cement was used for
0.8% of the patients in either group.

/ 9c5b$$no30

1249

1250

Berbari et al.

Table 3. Microbiological findings for 462 cases with prosthetic joint

infection seen at the Mayo Clinic between 1969 and 1991.
Finding

No. (%)

Staphylococcus aureus
Polymicrobial etiology
Coagulase-negative staphylococci
Negative culture
Streptococci
Gram-negative bacilli
Anaerobes
Others*

101
88
86
57
42
38
29
21

(22)
(19)
(19)
(12)
(9)
(8)
(6)
(5)

* Enterococcus species (6), Corynebacterium species (3), Listeria monocytogenes (1), unknown (5), Mycobacterium tuberculosis (3), Candida albicans
(1), Brucella suis (1), and Geotrichum species (1).

/ 9c5b$$no30

10-14-98 12:06:56

The median duration of diabetes mellitus before arthroplasty

was 4.5 years (range, 128 years) for the cases and 5.5 years
(range, 114 years) for the controls. No significant difference in
the number of diabetics who required insulin was noted between
the two groups (eight cases [27.6%] vs. four controls [28.6%]).
Native joint septic arthritis occurred a median of 29.5 years
(range, 1 56 years) before index arthroplasty for the cases
and a median of 24.5 years (range, 2 56 years) before index
arthroplasty for the controls. The microbiological etiology of
native joint septic arthritis was known for 36.4% of cases
(S. aureus [7 cases], Mycobacterium tuberculosis [3], streptococci [1], and polymicrobial [1]) and 35.7% of controls (S.
aureus [2 controls], streptococci [2], and M. tuberculosis [1]).
None of the cases or controls had gross, pathological, or microbiological evidence of septic arthritis at the time of index arthroplasty.
The distribution of the National Nosocomial Infections Surveillance (NNIS) System surgical patient risk index scores [29] was
as follows: 0190 cases (41%) and 260 controls (57%); 1
208 cases (45%) and 181 controls (39%); and 260 cases (13%)
and 16 controls (4%). The distribution of the American Society
for Anesthesia preoperative assessment score [32] was as follows:
I31 cases (7%) and 48 controls (10%); II219 cases (47%)
and 281 controls (61%); III200 cases (43%) and 120 controls
(26%); and IV9 cases (2%) and 8 controls (2%). The surgery
time was 3 hours for 119 cases (25.8%) and 89 controls (19.3%).
An additional surgical procedure was performed during the
index hospitalization for 103 cases (22.3%) and 62 controls
(13.4%). These procedures were performed to treat postoperative
wound complications (including surgical site infection, wound
drainage, wound hematoma, or wound dehiscence) in 29 cases
(28%) and one control (1.6%). Surgical procedures unrelated to
postoperative wound complications (such as total joint arthroplasty on a joint other than the index joint), other musculoskeletal
surgical procedures, and other surgical procedures were performed
for 74 cases (72%) and 61 controls (98.4%).
Surgical site infections that did not involve the joint prosthesis developed during the index hospitalization a median of 13
days (range, 3 209 days) and 12.5 days (range, 11 16 days)
after the time of index arthroplasty for the cases and controls,
respectively. The microbiological etiologies of the surgical site
infections in 58 cases were as follows: polymicrobial (28),
S. aureus (6), gram-negative bacilli (6), coagulase-negative
staphylococci (4), gram-positive bacilli (3), streptococci (1),
anaerobes (1), and unknown (9). The microbiological etiologies
of the surgical site infections in four controls were as follows:
S. aureus (2), polymicrobial (1), and coagulase-negative staphylococci (1). Fourteen cases (24.1%) and none of the controls
with surgical site infections required surgical treatment.
Multivariate Analysis

A multiple logistic regression model indicated that the two

best predictors of prosthetic joint infection in cases compared
with controls were the development of a surgical site infection

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

years (range, 1 39 years) for the controls. Forty-six (45.1%)

of the cases with rheumatoid arthritis and 29 (40.6%) of the
controls with rheumatoid arthritis required steroid therapy for
management of their disease.
Reasons for steroid therapy for the cases were rheumatoid
arthritis (46), other inflammatory arthritides (11), chronic skin
disease (3), chronic liver disease (2), asthma (1), myasthenia
gravis (1), renal transplant (1), malignancy (1), and unknown
(8). Reasons for steroid therapy for the controls were rheumatoid arthritis (28), other inflammatory arthritides (3), chronic
obstructive pulmonary disease (2), renal transplant (1), CNS
vasculitis (1), chronic relapsing neuropathy (1), and unknown
(2). The median daily dosage of prednisone for the 63 cases
(85.1%) for whom data were available was 6 mg/d (range, 2.5
62 mg/d). The median daily dosage of prednisone for the 37
controls (95%) 37 was 5 mg/d for whom data were available
(range, 2 20 mg/d). The median duration of therapy for the
72 cases (97.3%) for whom data were available was 104 weeks
(range, 1 1,560 weeks). The median duration of therapy for
the 37 controls was 312 weeks (range, 4 1,612 weeks).
Malignancy was diagnosed within 5 years of index arthroplasty for 33 cases (7.1%) (prostate [5], gastrointestinal [3],
breast [5], gynecologic [6], renal [1], osteosarcoma [3], other
bone tumors [4], bladder cancer [1], squamous cell carcinoma
[1], metastatic disease of unknown origin [2], Sezary syndrome
[1], and lymphoproliferative disorder [1]) and 13 controls
(2.8%) (prostate [4], other bone tumors [2], gastrointestinal [1],
breast [1], kidney [1], parotid [1], testicular [1], metastatic
carcinoid [1], and ovarian [1]). Ten cases had an underlying
primary joint tumor or metastasis before arthroplasty (metastasis [5], osteosarcoma [3], and giant-cell tumor [2]), whereas
only two controls had an underlying joint tumor (chondrosarcoma [1] and lymphoma [1]). Fifteen cases (45.4%) and three
controls (23.1%) for whom malignancy was diagnosed within
5 years of arthroplasty were believed to have a malignancy
at the time of index arthroplasty. None of the patients were
neutropenic (absolute neutrophil count, 500/mm3) at the time
of arthroplasty.

CID 1998;27 (November)

CID 1998;27 (November)

Risk Factors for Prosthetic Joint Infection

1251

Table 4. Univariate analysis of risk factors for prosthetic joint infection in cases with prosthetic joint infection and their matched controls
who were seen at the Mayo Clinic between 1969 and 1991.
Cases
(n 462)

Risk factor

Matched OR
(95% CI)

P value

172
4
1
4
295
102
74
33
10
29
33

(39.4)
(0.9)
(0.2)
(0.9)
(96.1)
(22.1)
(16)
(7.1)
(2.2)
(6.3)
(7.1)

182
2
1
7
292
69
39
13
2
14
14

(40.6)
(0.4)
(0.2)
(1.5)
(79.8)
(15)
(8.4)
(2.8)
(0.4)
(3.0)
(3.0)

0.9 (0.7 1.2)

1.9 (0.3 10.4)
1.4 (0.1 25.0)
0.6 (0.2 1.9)
1.2 (0.6 2.4)
2 (1.3 3.0)
2.0 (1.3 3.1)
2.4 (1.3 4.7)
4.1 (0.9 19.4)
2.3 (1.2 4.8)
2.3 (1.2 4.3)

NS
NS
NS
NS
NS
.01
.01
.01
.1
.05
.01

19
79
121
102
14
44

(12.2)
(29)
(26.2)
(22.1)
(18.7)
(10.6)

11
53
67
91
9
44

(7)
(21.4)
(14.5)
(19.7)
(16.4)
(10.6)

2.2
2.0
2.2
1.1
2.9
1.0

(0.8 5.9)
(1.2 3.2)
(1.5 3.1)
(0.8 1.6)
(0.2 41.0)
(0.6 1.6)

NS
.01
.01
NS
NS
NS

208 (45.4)
60 (13.1)
4.3 (1 37)
407 (88.1)
433 (93.7)
132 (28.6)
4 (0.9)

181 (40)
16 (3.5)
3.6 (1 85)
401 (86.8)
437 (94.6)
108 (23.4)
7 (1.5)

1.7
5.2
1.0
1.1
0.8
1.3
0.6

(1.3 2.3)
(2.8 9.5)
(1.0 1.1)
(0.8 1.7)
(0.4 1.6)
(1.0 1.8)
(0.2 2.0)

.01
.01
.01
NS
NS
.1
NS

356
10
44
187

331
12
52
212

1.5
0.8
0.8
0.8

(1.0 2.2)
(0.3 2.0)
(0.4 1.3)
(0.6 1.05)

.05
NS
NS
NS

(77.1)
(2.2)
(9.5)
(40.5)

29.9 (8 428)
58 (12.5)
148 (32.0)
72 (15.6)
23 (5)
453 (98.0)
2.4 (1 6)
50 (10.8)
21 (4.5)
22 (4.8)
17 (3.7)
123 (26.6)

(71.6)
(2.6)
(11.2)
(45.9)

21.8 (6 121)
4 (0.9)
87 (18.8)
30 (6.5)
1 (0.2)
459 (99.3)
2.3 (1 6)
36 (7.8)
11 (2.4)
9 (1.9)
8 (1.7)
62 (13.4)

1.03 (1.0 1.04)

27.5 (6.7 112.8)
2.1 (1.5 2.9)
2.7 (1.7 4.3)
21.5 (2.9 160.0)
0.3 (0.1 1.2)
1.2 (0.95 1.5)
1.4 (0.9 2.2)
1.8 (0.9 3.8)
3.6 (1.3 9.7)
2.3 (0.9 5.4)
2 (1.3 2.8)

.01
.01
.01
.01
.01
.1
.1
NS
.1
.01
.1
.01

NOTE. Unless stated otherwise, data are no. (%) with indicated risk factor. NNIS National Nosocomial Infections Surveillance; NS not significant.
* Number of cases for whom risk factor data were available.

Number of controls for whom risk factor data were available.

Primary or metastatic malignancy involving the site of the index arthroplasty.

Pneumonia, bloodstream infection, gastrointestinal infection, and other infections.

not involving the prosthesis (OR, 35.9; 95% CI, 8.3 154.6)
and an NNIS System surgical patient risk index score of 1
(OR, 1.7; 95% CI, 1.2 2.3) or 2 (OR, 3.9; 95% CI, 2.0 7.5);
other risk factors were a systemic malignancy and prior joint

/ 9c5b$$no30

10-14-98 12:06:56

arthroplasty (table 5). There were no significant interactions

between the matched variables and the main effect variables
or between any of the main effect variables that were significant
in the multivariate model.

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

Host risk factor

Obesity (436*/448)
Cirrhosis
Hemophilia
Psoriasis
Creatinine clearance, 30 mL/min (307/366)
Rheumatoid arthritis
Steroid therapy
Malignancy
Joint malignancy
Diabetes mellitus
Prior native joint septic arthritis
Malnutrition
Albumin level, 34 g/L (156/158)
Absolute lymphocyte count, 1.5 1 109/L (272/248)
Prior arthroplasty
Prior joint surgery
Preoperative bacteriuria (75/55)
Preoperative pyuria (416/413)
Index arthroplasty risk factor
NNIS System surgical patient risk index score (458/457)
1 vs. 0
2 vs. 0
Mean time from admission to arthroplasty in d (range)
Staff surgeon
Cemented arthroplasty vs. uncemented arthroplasty
P.M. surgery
Other surgery at same time as index arthroplasty
Blood transfusion
Homologous
Autologous
Laminar airflow
Antimicrobial prophylaxis 120 min before surgery
Postoperative risk factors
Mean duration of index hospitalization in d (range)
Surgical site infection
Wound drainage
Wound hematoma
Wound dehiscence
Surgical drain
Mean duration of surgical drain in d (range) (425/414)
Foley urinary catheter
Nosocomial urinary tract infection
Pressure sores or decubitus ulcer
Nosocomial infection
Other surgery during index hospital admission

Controls
(n 462)

1252

Berbari et al.

Table 5. Significant risk factors in a multivariate analysis of risk

factors for prosthetic joint infection in 462 cases with prosthetic joint
infection and their matched controls who were seen at the Mayo
Clinic between 1969 and 1991.

Risk factor
Postoperative surgical site
infection
NNIS System surgical patient risk
index score
1 vs. 0
2 vs. 0
Systemic malignancy
Prior joint arthroplasty

Matched OR
(95% CI)

P value

35.9 (8.3 154.6)

.01

1.7
3.9
3.1
2.0

(1.2 2.3)
(2.0 7.5)
(1.3 7.2)
(1.4 3.0)

.05
.01
.01
.01

NOTE. NNIS National Nosocomial Infections Surveillance.

This large case-control study defined four independent risk

factors for prosthetic joint infection by multivariate analysis:
the development of a surgical site infection not involving the
prosthesis, an NNIS System surgical patient risk index score
of 1, a history of malignancy, or a history of prior total joint
arthroplasty. Strengths of this study include strict definition of
cases, controls, and potential risk factors as well as a study
design that controlled for different clinical practices occurring
during the study period and differences in follow-up between
cases and controls. Furthermore, to avoid reporting bias between cases and controls, only the risk factors present before
or during the index joint arthroplasty and subsequent hospitalization were reviewed.
The NNIS System surgical patient risk index score for identifying patients at high risk for postoperative surgical site infection has been shown to be a better predictor of surgical site
infection than individual components of the index [29, 30, 33,
34]. This index includes measurements of intrinsic patient risk
[37], surgical wound classification, and duration of the operative procedure. The NNIS System surgical patient risk index
consists of scoring each operation by adding the number of
risk factors among these three measurements. One point is
given if the American Society of Anesthesiologists preoperative assessment score is 3, 4, or 5; another point is given if the
operation time is 3 hours for prosthetic joint arthroplasty,
and a third point is given if the wound classification is 3 or 4
[33]. It is a modification of another index developed in the
study on the efficacy of nosocomial infection control (SENIC)
[38, 39].
For 5,696 patients undergoing THA or TKA at 44 centers,
the NNIS System surgical patient risk index was effective in
identifying patients at high risk for both superficial surgical
site infection and prosthetic joint infection [33]. Prosthetic joint
infection alone was not separately analyzed. In this study, an
NNIS System surgical patient risk index score of 1 was a

/ 9c5b$$no30

10-14-98 12:06:56

significant risk factor for the development of prosthetic joint

infection even after adjustment for the development of superficial surgical site infection. We believe that the value of the
NNIS System surgical patient risk index to health care providers and patients is in the simplicity of its calculation and the
ready availability of the necessary information for all patients
undergoing prosthetic joint implantation.
To our knowledge, this is the first study to observe an association between a diagnosis of malignancy not involving the
index joint and the development of prosthetic joint infection.
The absence of an association in prior studies may be due to
a lack of statistical power [4, 14]. Possibilities for the association of malignancy and prosthetic joint infection include the
immunosuppressive effects of treatment for malignancy that
are unrelated to neutropenia or steroid therapy and simply unknown factors associated with the malignancy itself [40]. The
specific reasons for increased risk in patients with a malignancy
require further elucidation.
Surgical site infection not involving the prosthesis has previously been shown to be a significant risk factor for the development of prosthetic joint infection [16, 18, 41 44]. In a carefully analyzed prospective study [18], although multiple
indicators of poor wound healing were associated with an increased risk of joint infection in univariate analyses, only superficial wound infection (defined as erythema 1 cm from the
incision and an unhealed wound at discharge) were independently associated with an increased risk of deep infection in
multivariate analyses. Our study confirms these results with
use of the CDC definition of superficial surgical site infection
[31].
A history of joint arthroplasty on the index joint has been
consistently recognized as a risk factor for prosthetic joint
infection in various retrospective cohort studies [17, 18, 43,
45]. In a review of 4,240 large joint arthroplasties followed up
for a mean of 2.5 to 3.5 years, patients undergoing revision
arthroplasty were eight times more likely to have a prosthetic
joint infection than were patients with primary arthroplasty.
Other investigators have found that the risk of prosthetic joint
infection increases with the number of previous joint arthroplasties [45]. It has been unclear from these studies whether
the increased risk of prosthetic joint infection in patients with
prior total joint arthroplasties is due to an increased number of
comorbid conditions, a prolonged operating time, an increased
number of blood transfusions, or a higher frequency of postoperative wound complications. In this study, prior joint arthroplasty on the index joint remained a significant risk factor even
after accounting for these other variables (suggesting that other
factors such as local tissue scarring or increased dead space
are possible etiologies).
The increased risk of prosthetic joint infection in patients
with rheumatoid arthritis that was observed in the univariate
analysis is consistent with previously reported findings [4, 8,
14, 15, 18, 43, 46]. In a univariate analysis of 53 prosthetic
joint infections that developed after 4,240 procedures followed

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

Discussion

CID 1998;27 (November)

CID 1998;27 (November)

Risk Factors for Prosthetic Joint Infection

References
1. Graves EJ. Detailed diagnoses and procedures, National Hospital Discharge Survey, 1993. Vital Health Stat 13 1995; 122:1 288.
2. Rand JA, Morrey BF, Bryan RS. Management of the infected total joint
arthroplasty. Orthop Clin North Am 1984; 15:491 504.
3. Charnley J. Postoperative infection after total hip replacement with special
reference to air contamination in the operating room. Clin Orthop 1972;
87:167 87.

/ 9c5b$$no30

10-14-98 12:06:56

4. Wilson MG, Kelley K, Thornhill TS. Infection as a complication of total

knee replacement arthroplasty. Risk factors and treatment in sixtyseven cases. J Bone Joint Surg Am 1990; 72:878 83.
5. Schmalzried TP, Amstutz HC, Au MK, Dorey FJ. Etiology of deep sepsis
in total hip arthroplasty. The significance of hematogenous and recurrent
infections. Clin Orthop 1992; 280:200 7.
6. Bengtson S, Knutson K. The infected knee arthroplasty. A 6-year followup of 357 cases. Acta Orthop Scand 1991; 62:301 11.
7. Lidwell OM, Lowbury EJ, Whyte W, Blowers R, Stanley SJ, Lowe D.
Effect of ultraclean air in operating rooms on deep sepsis in the joint
after total hip or knee replacement: a randomised study. BMJ 1982;
285:10 4.
8. Salvati EA, Robinson RP, Zeno SM, Koslin BL, Brause BD, Wilson
PD Jr. Infection rates after 3175 total hip and total knee replacements
performed with and without a horizontal unidirectional filtered air flow
system. J Bone Joint Surg Am 1982; 64:525 35.
9. Grogan TJ, Dorey F, Rollins J, Amstutz HC. Deep sepsis following total
knee arthroplasty. Ten-year experience at the University of California
at Los Angeles Medical Center. J Bone Joint Surg Am 1986; 68:
226 34.
10. Powers KA, Terpenning MS, Voice RA, Kauffman C. Prosthetic joint
infections in the elderly. Am J Med 1990; 88(5N):9N 13N.
11. Ahlberg A, Carlsson AS, Lindgren L. Hematogenous infection in total
joint replacement. Clin Orthop 1978; 137:69 75.
12. Sculco TP. The economic impact of infected joint arthroplasty. Orthopedics 1995; 18:871 3.
13. Hanssen AD, Osmon DR, Nelson CL. Prevention of deep periprosthetic
joint infection. Instr Course Lect 1997; 46:555 67.
14. Fitzgerald RH Jr, Nolan DR, Ilstrup DM, Van Scoy RE, Washington JA
II, Coventry MB. Deep wound sepsis following total hip arthroplasty.
J Bone Joint Surg Am 1977; 59:847 55.
15. Insall J, Scott WN, Ranawat CS. The total condylar knee prosthesis. A
report of two hundred and twenty cases. J Bone Joint Surg Am 1979;
61:173 80.
16. Petty W, Bryan RS, Coventry MB, Peterson LF. Infection after total knee
arthroplasty. Orthop Clin North Am 1975; 6:1005 14.
17. Rand JA, Fitzgerald RH Jr. Diagnosis and management of the infected
total knee arthroplasty. Orthop Clin North Am 1989; 20:201 10.
18. Wymenga AB, van Horn JR, Theeuwes A, Muytjens HL, Slooff TJ. Perioperative factors associated with septic arthritis after arthroplasty. Prospective multicenter study of 362 knee and 2651 hip operations. Acta Orthop
Scand 1992; 63:665 71.
19. Ilsrup DM. Statistical considerations and the science of surveillance. In:
Morrey BF, ed. Reconstructive surgery of the joints. 2nd ed. Vol 2.
New York: Churchill-Livingstone, 1996:225 30.
20. Kurland LT, Molgaard CA. The patient record in epidemiology. Sci Am
1981; 245:54 63.
21. Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism
Association 1987 revised criteria for the classification of rheumatoid
arthritis. Arthritis Rheum 1988; 31:315 24.
22. National Diabetes Data Group. Classification and diagnosis of diabetes
mellitus and other categories of glucose intolerance. Diabetes 1979;
28:1039.
23. Cockcroft DW, Gault MH. Prediction of creatinine cleareance from serum
creatinine. Nephron 1976; 16:31 41.
24. Brumfitt W. Urinary cell counts and their value. J Clin Pathol 1965; 18:
550 5.
25. Hirsch J. NIH consensus conferences: by whom, for what? Obesity. J Nutr
1986; 116:1854 6.
26. Greene KA, Wilde AH, Stulberg BN. Preoperative nutritional status of
total joint patients. Relationship to postoperative wound complications.
J Arthroplasty 1991; 6:321 5.
27. Anderson CF, Moxness K, Meister J, Burritt MF. The sensitivity and
specificity of nutrition-related variables in relationship to the duration

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

up for a mean of 2.5 to 3.5 years, the risk of prosthetic joint

infection following THA and TKA was 1.8- and 2.4-fold times
higher, respectively, in patients with rheumatoid arthritis than
in patients with degenerative joint disease [43]. Similarly, in
a study of 67 patients who developed TKA infection after 4,171
TKA implantations between 1973 and 1980 [4], the incidence
of prosthetic joint infection among patients with rheumatoid
arthritis was 2.5-fold higher than that among patients with
degenerative joint disease.
Whether the increased risk of prosthetic joint infection in
patients with rheumatoid arthritis is due to an increased incidence of early wound infection [47], a higher prevalence of
revision surgeries, skin ulcers, steroid therapy, or immunologic
abnormalities associated with rheumatoid arthritis has remained
unclear [46 49]. In our study, rheumatoid arthritis was not a
significant risk factor for the development of prosthetic joint
infection after adjustment for these other factors, which suggests that these variables are the reason for the increased risk of
infection rather than the immunologic abnormalities associated
with rheumatoid arthritis.
There are at least two limitations to this study. It is possible
that risk factors that occur infrequently may not have been
identified. It is also possible, given the retrospective study
design, that the attending physician may have unintentionally
misclassified prosthetic joint infection as a surgical site infection not involving the prosthesis, therefore biasing the results
of this study. Although this misclassification may have occurred and the true risk of superficial surgical site infection
may be less than what we have estimated, we believe that
these data are informative as they accurately reflect the risk of
subsequent prosthetic joint infection in patients for whom a
superficial surgical site infection is diagnosed by an attending
physician.
This study has important implications for patients and health
care providers caring for patients with prosthetic joints and for
future investigation of the prevention of deep prosthetic joint
infection. It provides the patient and the health care provider
with information that allows improved preoperative risk assessment and counseling regarding the risk of prosthetic joint infection. It also identifies patient populations for whom physicians
should have an increased index of suspicion for difficult-todiagnose prosthetic joint infection. Last, these data identify
high-risk patients for whom additional prophylaxis measures
may be required. Future studies regarding prevention of infection should focus on these high-risk patients.

1253

1254

28.

29.

30.

31.
32.
33.

35.
36.
37.
38.

of hospital stay and the rate of complications. Mayo Clin Proc 1984;
59:477 83.
Anderson CF, Wochos DN. The utility of serum albumin values in the
nutritional assessment of hospitalized patients. Mayo Clin Proc 1982;
57:181 4.
National Nosocomial Infections Surveillance (NNIS) report, data summary
from October 1986 April 1996, issued May 1996. Am J Infect Control
1996; 24:380 8.
Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification of
CDC definitions of surgical wound infections. Am J Infect Control
1992; 20:271 4.
Garner JS, Jarvis WR, Emori TG, Horan TC, Hughes JM. CDC definitions
for nosocomial infections, 1988. Am J Infect Control 1988; 16:128 40.
New classification of physical status. Anesthesiology 1963; 24:111.
Culver DH, Horan TC, Gaynes RP, et al. Surgical wound infection rates
by wound class, operative procedure, and patient risk index. National
Nosocomial Infections Surveillance System. Am J Med 1991; 91(3B):
152S 7S.
Horan TC, Culver DH, Gaynes RP, Jarvis WR, Edwards JR, Reid CR.
Nosocomial infections in surgical patients in the United States, January
1986 June 1992. National Nosocomial Infections Surveillance (NNIS)
System. Infect Control Hosp Epidemiol 1993; 14:73 80.
Schlesselman J, ed. Case-control studies. Design, conduct, analysis. New
York: Oxford University Press, 1982.
Efron B, Tibhsirani R. An introduction to the bootstrap. New York: Chapman & Hall, 1993.
Haley RW. Nosocomial infections in surgical patients: developing valid
measures of intrinsic patient risk. Am J Med 1991; 91(3B):145S 51S.
Haley RW, Morgan WM, Culver DH, et al. Update from the SENIC
project. Am J Infect Control 1985; 121:182 205.

/ 9c5b$$no30

10-14-98 12:06:56

CID 1998;27 (November)

39. Haley RW, Culver DH, Morgan WM, White JW, Emori TG, Hooton TM.
Identifying patients at high risk of surgical wound infection. A simple
multivariate index of patient susceptibility and wound contamination.
Am J Epidemiol 1985; 121:206 15.
40. Smith RT. Cancer and the immune system. Pediatr Clin North Am 1994;
41:841 50.
41. Glynn K, Sheehan MJ. An analysis of the causes of total hip and total
knee arthroplasty. Clin Orthop 1983; 178:202 6.
42. Surin VV, Sundholm K, Bachman L. Infection after total hip replacement.
With special reference to discharge from the wound. J Bone Joint Surg
Br 1983; 65:412 8.
43. Poss R, Thornhill TS, Ewald FC, Thomas WH, Batte NJ, Sledge CB.
Factors influencing the incidence and outcome of infection following
total joint arthroplasty. Clin Orthop 1984; 182:117 26.
44. Bengston S, Knutson K, Lidgren L. Treatment of infected knee arthroplasty. Clin Orthop 1989; 245:173 8.
45. Ahnfelt L, Herberts P, Malchau H, Andersson GB. Prognosis of total hip
replacement. A Swedish multicenter study of 4,664 revisions. Acta
Orthop Scand Suppl 1990; 238:1 26.
46. Goldenberg DL. Infectious arthritis complicating rheumatoid arthritis
and other chronic rheumatic disorders. Arthritis Rheum 1989; 32:
496 501.
47. White RH, McCurdy SA, Marder RA. Early morbidity after total hip
replacement: rheumatoid arthritis versus osteoarthritis. J Gen Intern Med
1990; 5:304 9.
48. Wilton JM, Gibson T, Chuck CM. Defective phagocytosis by synovial fluid
and blood polymorphonuclear leucocytes in patients with rheumatoid
arthritis. I. The nature of the defect. Rheumatol Rehabil 1978; (Suppl):
25 36.
49. Turner RA, Schumacher R, Myers AR. Phagocytic function of polymorphonuclear leukocytes in rheumatic diseases. J Clin Invest 1973; 52:
1632 5.

cida

UC: CID

Downloaded from http://cid.oxfordjournals.org/ by guest on December 31, 2015

34.

Berbari et al.