Delayed Wound Closure Increases Deep-Infection

Rate Associated with Lower-Grade Open Fractures

A Propensity-Matched Cohort Study
Richard J. Jenkinson, MD, MSc, FRCS(C), Alexander Kiss, PhD, Samuel Johnson, MD,
David J.G. Stephen, MD, FRCS(C), and Hans J. Kreder, MD, MPH, FRCS(C)
Investigation performed at the Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Ontario, Canada
Background: Primary closure of skin wounds after debridement of open fractures is controversial. The purpose of the
present study was to determine whether primary skin closure for grade-IIIA or lower-grade open extremity fractures is
associated with a lower deep-infection rate.
Methods: We identied 349 Gustilo-Anderson grade-I, II, or IIIA fractures treated at our level-I academic trauma center
from2003 to 2007. Eighty-seven injuries were treated with delayed primary closure, and 262 were treated with immediate
closure after surgical debridement. After application of a propensity score-matching algorithm to balance prognostic
factors, 146 open fractures (seventy-three matched pairs) were analyzed.
Results: After application of a propensity score-matching algorithm with adjustment for age, sex, time to debridement,
American Society of Anesthesiologists (ASA) class, fracture grade, evidence of gross contamination, and a tibial fracture
rather than a fracture at another anatomic site, the two treatment groups were compared with respect to the prevalence of
infection. Deep infection developed at the sites of three of the seventy-three fractures treated with immediate closure
(infection rate, 4.1%; 95% condence interval [CI], 0.86 to 11.5) compared with thirteen in the matched group of seventy-
three fractures treated with delayed primary closure (infection rate, 17.8%; 95%CI, 9.8to 28.5) (McNemar test, p =0.0001).
Conclusions: Immediate closure of carefully selected wounds by experienced surgeons treating grade-I, II, and IIIA open
fractures is safe and is associated with a lower infection rate compared with delayed primary closure.
Level of Evidence: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.
T
reatment standards for open fractures require timely
irrigation and adequate debridement
1
. Traumatic open-
fracture wounds traditionally have been left open after
the initial debridement in order to minimize the risk of later
deep infection, especially with Clostridium
2
. This treatment
strategy is traced to the experience of trauma surgeons fromthe
pre-antibiotic era, especially during World War I and World
War II
3,4
. Delayed wound closure until a few days post-injury is
currently advocated by many surgeons to allow drainage of any
collecting infectious material and to allow for a universal second-
look debridement
2,5
. However, with advances in stabilization
methods, antibiotics, and wound management, the strict avoid-
ance of immediate wound closure has beenchallenged, with several
investigators reporting low infection rates
6-8
. Immediate wound
closure after initial debridement has the advantage of providing
immediate soft-tissue cover to the traumatized limb as well as some
protection against nosocomial pathogens
9
. Also, subsequent visits
to the operating room for second-look debridements may be
avoided if immediate closure is chosen, thereby simplifying and
streamlining management of the traumatized patient.
There are no clear guidelines for determining the time
frame for safe closure of traumatic open fracture wounds, and
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any
aspect of this work. One or more of the authors, or his or her institution, has had a nancial relationship, in the thirty-six months prior to submission of this
work, with an entity in the biomedical arena that could be perceived to inuence or have the potential to inuence what is written in this work. No author has
had any other relationships, or has engaged in any other activities, that could be perceived to inuence or have the potential to inuence what is written in this
work. The complete Disclosures of Potential Conicts of Interest submitted by authors are always provided with the online version of the article.
Peer Review: This article was reviewed by the Editor-in-Chief and one Deputy Editor, and it underwent blinded review by two or more outside experts. It was also reviewed
by an expert in methodology and statistics. The Deputy Editor reviewed each revision of the article, and it underwent a nal reviewby the Editor-in-Chief prior to publication.
Final corrections and clarications occurred during one or more exchanges between the author(s) and copyeditors.
380
COPYRIGHT 2014 BY THE JOURNAL OF BONE AND JOINT SURGERY, INCORPORATED
J Bone Joint Surg Am. 2014;96:380-6
d
http://dx.doi.org/10.2106/JBJS.L.00545
this determination is left to the judgment and experience of the
treating orthopaedic surgeon. Higher-risk open fractures would
be expected to have higher rates of more conservative wound
management. We compared the rate of subsequent deep infec-
tion between patients in whom an open fracture had been
treated with immediate primary closure after debridement and
those in whom the fracture had been treated with delayed
primary closure. Bias was reduced with use of a matched-pairs
design. We hypothesized that early wound closure would reduce
the rate of subsequent deep infection.
Materials and Methods
C
onsecutive patients treated for an open extremity fracture from January 1,
2003, to January 1, 2007, at our level-I trauma center were identied with
use of our trauma database. Open hand and pelvic fractures were excluded. We
supplemented case identication with the billing database of the central or-
thopaedic department with use of specic coding identiers for open fractures
(E556 modier). Medical records were abstracted by two orthopaedic surgeons
(R.J.J. and S.J.). We identied 417 patients who had a total of 459 fractures. The
inclusion and exclusion criteria are shown in Figure 1. Gustilo-Anderson grade-
IIIB fractures usually are not amenable to primary closure and were excluded.
Gustilo-Anderson grade-IIIC fractures also were excluded because these in-
juries often are treated simultaneously with fasciotomies, for which skin closure
is contraindicated. This left 345 patients with a total of 415 fractures. Thirteen
patients died from their traumatic injuries during the index hospital stay and
were excluded, which left 332 patients available for follow-up. Complete
follow-up was dened as twelve months, which was not achieved for thirty-
eight patients (rate of complete follow-up = 89%). Our nal cohort consisted of
294 patients with a total of 349 Gustilo-Anderson grade-I, II, or IIIA fractures.
The collected patient demographic characteristics included age, sex, and
American Society of Anesthesiologists (ASA) class. The collected injury vari-
ables included the injury severity score (ISS), evidence of gross contamination
of the fracture, the anatomic fracture location, and the time delay to surgery
(dened as from ambulance call time to surgical start time). AGustilo-Anderson
grade
1,10
was assigned to the fracture on the basis of the surgical description of the
injury after debridement. Gross contaminationwas documented if dirt or foreign
material was present. None of the wounds had fecal or farmyard contamination
or contamination with grass. The primary outcome measure for the present study
was deep infection, dened as infection of the injured bone and deep tissue
necessitating an unplanned operative irrigation and debridement at more than
two weeks after the injury. We chose a two-week cutoff in order to address the
concern of whether an early debridement was planned or unplanned. Planned
repeat debridements and supercial infections not requiring surgery were not
considered to be deep infections. All deep infections were treated with one or
more surgical debridements, possible implant removal, and/or skeletal stabili-
zation. Initial characteristics of the injuries are shown in Table I.
The standard treatment protocol at our institution included intravenous
antibiotics immediately administered on arrival in the emergency department
and continued during hospitalization until at least twenty-four hours after de-
nitive wound closure. Intravenous cefazolin was administered, although clin-
damycin was used when a severe penicillin allergy was known. Gentamicin was
added for grade-III open fractures. Debridements were performed urgently,
TABLE I Baseline Characteristics
Variable
Average age (yr) 40.7
Male sex* 240 (68.8%)
ASA class >2* 66 (18.9%)
ISS >25 points* 155 (44.4%)
Gunshot mechanism* 28 (8.0%)
Average injury to debridement time (hr) 11.27
Average injury to rst antibiotic time (hr) 2.53
Tibial fracture* 148 (42.4%)
Gross contamination* 140 (40.1%)
Fracture grade*
Grade I 53 (15.2%)
Grade II 141 (40.4%)
Grade IIIA 155 (44.4%)
Primary closure* 262 (75.0%)
Deep infection* 25 (7.2%)
*Values are given as the number of patients, with the percentage
in parentheses.
Fig. 1
Patients included and excluded from the study.
381
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES
on the basis of availability of the operating room. Normal saline solution was
used for irrigation, with gravity or pulse lavage used at the discretion of the
treating surgeon. Wound culture specimens were not routinely taken at the
time of initial debridement. Antibiotic choice, xation method, and wound
closure were also at the discretion of the treating physicians. Vacuum-assisted
closure dressings were not used for any of these fractures. Second-look de-
bridement after approximately forty-eight hours was performed routinely
when delayed closure was chosen and was performed in patients who had
undergone primary closure when chosen by the treating surgeon on the basis
of the impression of the adequacy of the debridement.
All statistical analyses were performed with use of SAS version 9.2 (SAS
Institute, Cary, North Carolina) with input from a statistician (A.K.). To adjust
for confounding by indication
11
, a propensity-score
12,13
matched-cohort study
was developed from the original data set (Table I). Injury characteristics were
used in a logistic regression model to predict the likelihood of the need for
treatment with delayed wound closure. As eighty-seven patients were managed
with delayed wound closure, up to eight degrees of freedom could be specied
in the propensity score
14
. Dichotomous and continuous variables used one
degree of freedom each, whereas the three-level variable used two degrees of
freedom. The factors considered to be the most important confounders also
contributing to deep-infection risk were chosen for the propensity-score al-
gorithm. These factors included patient age, sex, time delay to debridement,
fracture grade (Gustilo-Anderson grade I, II, or IIIA), evidence of gross con-
tamination, tibial compared with nontibial site, and ASAclass (1 or 2 compared
with 3 or higher). These factors were chosen, based on consensus among the
investigators, as the factors most important for predicting later infection but
also as those most divergent between the immediate and delayed-closure groups
(Table II). A one-to-one matching algorithm
15
was used to pair injuries with a
similar propensity for delayed wound closure. This algorithm searches for the
most exact match available to eight decimal places and then works back to one
TABLE II Characteristics of Wound Closure Treatment Groups Prior to Matching
Variable Primary Closure (N = 262) Delayed Closure (N = 87) P Value
Average age (yr) 41.3 39.7 0.24
Male sex 66.7% 74.7% 0.17
ASA class >2* 58 (22.1%) 8 (9.2%) 0.0016
Average injury to debridement time (hr) 12.06 8.90 0.0016
Average injury to rst antibiotic time (hr) 2.44 2.81 0.47
Tibial fracture* 104 (39.7%) 44 (50.6%) 0.0056
Gross contamination* 84 (32.1%) 56 (64.4%) 0.0001
Fracture grade*
Grade I 49 (18.7%) 4 (4.6%) 0.0001
Grade II 112 (42.7%) 29 (33.3%) 0.127
Grade IIIA 101 (38.5%) 54 (62.1%) 0.0001
Deep infection* 9 (3.4%) 16 (18.4%) 0.0008
*Values are given as the number of patients, with the percentage in parentheses.
TABLE III Characteristics of Wound Closure Treatment Groups After Matching
Variable Primary Closure (N = 73) Delayed Closure (N = 73) P Value
Average age (yr) 38.6 37.8 0.78
Male sex 76.7% 73.9% 0.70
ASA class >2* 7 (9.6%) 8 (11.0%) 0.78
Average injury to debridement time (hr) 9.75 8.8 0.30
Average injury to rst antibiotic time (hr) 1.92 2.75 0.09
Tibial fracture* 30 (41.1%) 33 (45.2%) 0.6191
Gross contamination* 44 (60.3%) 43 (58.9%) 0.8672
Fracture grade*
Grade I 4 (5.5%) 4 (5.5%) 1.0
Grade II 22 (30.1%) 27 (37.0%) 0.38
Grade IIIA 47 (64.4%) 42 (57.5%) 0.40
Deep infection* 3 (4.1%) 13 (17.8%) 0.0001
*Values are given as the number of patients, with the percentage in parentheses.
382
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES
decimal place. The maximum difference between propensity probabilities for
matching was set at 0.1. Seventy-three matched pairs of patients were identied.
Fourteen open fractures from the delayed-closure group were not paired because
of a lack of a suitable similar injury from the immediate primary-closure
group with which to match them. These unpaired injuries were more severe,
with higher open-fracture grades and more contamination. This process
generated matched pairs with similar injury characteristics for analysis of the
infection outcome.
The results were analyzed with standard descriptive statistics for the
baseline unpaired data. Infection rates were compared between the paired
cohorts with use of the McNemar test and conditional logistic regression.
Source of Funding
No external funding was received for this study.
Results
T
able II shows the fracture characteristics of the cohort of
patients prior to matching. As expected, patients managed
with delayed wound closure had fractures with more negative
prognostic factors, including a higher proportion of grade-IIIA
fractures (p = 0.0001), tibial fractures (p = 0.0056), and gross
contamination (p = 0.0001) (Fig. 2). The patients managed with
delayed wound closure also had a higher proportion of deep
infection before matching (18.4% compared with 3.4%, p =
0.0008), but this is an invalid comparison because of the se-
lection bias that results in more severe injuries being prefer-
entially treated with delayed wound closure.
Fig. 2
Comparison of major fracture characteristics between different closure treatment groups prior to matching.
Fig. 3
Comparison of major fracture characteristics between different closure groups after matching.
383
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES
After the propensity score-matching algorithmwas applied,
there were seventy-three matched pairs of patients and fractures
available for comparison (Table III). The two matched treatment
groups showed similar characteristics among all of the elements of
the propensity score, including fracture grade (p = 0.4), gross
contamination (p = 0.87), and tibial fractures (p = 0.62) (Fig. 3).
Patients with delayed closure had routine second-look debride-
ments, whereas planned second-look debridements were carried
out in fourteen (19%) of the seventy-three patients who had
undergone a primary closure.
Deep infection developed at the sites of three of the seventy-
three open fractures treated with primary closure (infection rate,
4.1%; 95% condence interval [CI], 0.86 to 11.5) compared with
thirteen of the seventy-three fractures treated with delayed closure
(infection rate, 17.8%; 95% CI, 9.8 to 28.5) (McNemar test,
p = 0.0001) (see Appendix). This nding suggests an absolute
risk reduction of 13.7% for development of deep infection for
the primary-closure group. This value corresponds to a number
needed to treat of 7.3 patients. Conditional logistic regression
was also used to conrm the signicance of this nding while
accounting for paired data. This yielded an odds ratio of 11.0
(95% CI, 1.42 to 85.2) times more likely to develop deep in-
fection if delayed rather than primary closure was performed.
No patient had a Clostridium infection. The infecting
organisms in each patient group are shown in Table IV.
Discussion
C
urrent orthopaedic literature does not provide clear direc-
tion to guide the choice between immediate wound closure
and the more traditional delayed techniques. A small random-
ized trial
6
showed no increase in the rate of infection with
primary closure. Primary closure has also been advocated by
DeLong et al.
8
, who reviewed 119 open fractures treated with
either primary closure or delayed methods. They found infection
rates to be independent of closure technique and suggested that
primary closure after thorough debridement is a viable option
when carried out by an experienced surgeon.
The deferral to surgical judgment is a common theme in
the literature regarding this subject. Most orthopaedic surgeons
would not suggest that primary closure be performed for patients
with persistent wound contamination with dirt, feces, and
nonviable tissue. Denitive closure of a wound without an ad-
equate debridement increases reoperation and infection rates.
However, objective characterization of what makes a wound
clean enough for primary closure is not available. Because
more severe open fractures will be considered not clean enough
for primary closure, a selection bias will occur in retrospective
studies on this topic, including the current study prior to
matching. There is confounding by indicationwhen the treatment
method is chosen at least partially on the basis of the external
prognostic factors that are associated with the injury
11
. In our
study, higher-grade tibial fractures with a degree of gross con-
tamination were more likely to be selected for delayed-closure
treatment. Because these more severe injuries carry a higher
risk for deep infection, we cannot conclude that treatment
with delayed closure leads to a higher likelihood of infection
without accounting for this confounding.
Traditional matching for each individual variable is often
too restrictive and does not allow enough pairs to be created
because of the constraints of nding an exact match for each
variable. The benet of a propensity score is that it allows
matching for many different factors by creating a composite
probability of receiving a particular treatmentin the present
study, either primary or delayed closure. Propensity-score match-
ing is a method of balancing treatment groups based on known
confounding variables and is particularly suited to account for
confounding by indication
12
. In the current study, the patients
managed with delayed closure were matched with patients who
had a similar severity of injury. After matching, similar groups of
patients and injuries were available for comparison. This creates
a pseudo-randomized or quasi-randomized study
12
in which the
two treatment groups can be effectively matched according to the
known prognostic factors that inform the propensity score.
However, this method can control only for known factors that
are included in the algorithm. Propensity score-matching tech-
niques are referred to as pseudo-randomized because matching of
treatment groups for all nontreatment variables is possible in
only a prospective, truly randomized study.
A higher infection rate was found among the fractures
treated with delayed closure, even after we accounted for im-
portant prognostic factors, including contamination, fracture
grade, and anatomic location. This nding suggests that
primary closure is safe and actually may be preferable to
delayed closure for selected lower-grade open fractures. In-
fection rates may be decreased when primary closure is em-
ployed for these carefully chosen injuries. The elimination of
TABLE IV Infecting Organisms by Closure Group*
Primary Closure Delayed Closure
Enterococcus faecalis MRSA
MRSA MRSA
MRSA Staphylococcus aureus
Staphylococcus aureus
Staphylococcus aureus
Coagulase-negative Staphylococcus
Pseudomonas aeruginosa
Pseudomonas aeruginosa
Enterococcus faecalis
Enterococcus faecalis
Escherichia coli
Gram-negative bacilli
Mixed aerobic organisms
Hafnia alvei
Candida
Negative culture in one patient
*MRSA = methicillin-resistant Staphylococcus aureus.
384
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES
an automatic second-look debridement for all open fractures
has the potential for streamlining patient care and for large
cost-savings.
The present study should be interpreted with its limita-
tions in mind. A propensity score is a valuable tool to use to
balance groups on the basis of known confounding variables.
A limitation of this technique is that unknown or nonmeasurable
variables cannot be accounted for. An experienced trauma surgeon
will use many different injury and patient characteristicsoften
unconsciously and intuitivelyto make an informed decision
about wound management. A randomized trial remains the gold
standard, as it allows balanced treatment groups that are based
on both known and unknown confounding variables. Addi-
tionally, the data used in the present study were collected ret-
rospectively. Complete initial and follow-up data were available
for 89% of the possible patients who could have been examined.
Efforts were made to dene the injury and outcome variables as
objectively as possible in order to minimize classication errors;
however, some inaccuracy is inevitable in a retrospective study.
Not all medical data were easily obtained via chart review; for
example, smoking status was inconsistently recorded and thus
was not a useful variable for data analysis. The ASA grade was
used as an indicator of medical frailty but does not provide a
complete picture of the patients medical status. Time to rst
antibiotic dose was similar between treatment groups (Table II).
We did not include this variable in the propensity-score algo-
rithmbecause we had to select the most important variables that
indicated a difference between treatment groups. Time to de-
bridement was included in the propensity-score algorithm be-
cause it was a major factor that differed betweentreatment groups.
This suggests that, for the surgeons at this institution, delay to
surgery likely was a consideration for choosing delayed rather
than primary closure. However, the data presented in this paper
do not directly support or refute delay to surgery as an important
factor in the development of infection after open fracture.
Dening a supercial infection is difcult with these ret-
rospective data, especially because many patients are given oral
antibiotics for various indications. Early antibiotic treatment
alone may suppress and delay but will not usually eliminate a
deep infection. This is why deep infection was chosen as our
primary outcome of interest. All deep infections were treated
with surgical debridement(s), possible implant removal, and/
or skeletal stabilization. Some infections that were excluded as
being supercial actually may have been deeper infections that
resolved without surgical debridement.
The Gustilo-Anderson classication scheme has limita-
tions interms of interobserver variability
16
; however, it is the most
frequently used classication scheme in current practice.
Although newer schemes are in development
17
, the Gustilo-
Anderson classication remains the most useful tool currently
available to characterize open-fracture severity. The chart ab-
stractors in our study both were experienced orthopaedic sur-
geons (an orthopaedic trauma fellowship-trained surgeon[R.J.J.]
and a trauma fellow with more than ve years of community
experience [S.J.]), which helped to improve the interpretation
of the clinical variables. Treatment was not standardized, and
individual surgeons made choices regarding primary closure
and many other treatment variables. Also, the follow-up time
to determine whether an infection was present was one year.
Posttraumatic infections can be relatively quiescent and may
not present for longer times. We acknowledge that a longer
duration of follow-up may result in more infections presenting
at a later follow-up interval. The patients were all treated at a
level-I academic trauma center. All participating clinicians had
an interest in orthopaedic trauma and would be expected to
have had sufcient clinical experience to perform adequate initial
debridements. This may help to explain why few deep infections
were found among patients managed with primary closure
without second-look debridement. Surgeons with less experience
in open fracture management may be more likely to perform
insufcient debridements, possibly leading to complications
from immediate closure. Also, the more severe (grade-IIIB and
grade-IIIC) injuries were excluded from the current study.
Generalization of the study ndings to more severe injuries and
those not treated with adequate initial debridement is not war-
ranted. It should also be stated that patients with fasciotomies or
those who are at high risk for developing compartment syn-
drome should not undergo primary wound closure.
Primary closure was chosen for several patients even
when a second-look debridement was planned, which may have
restored some degree of skin barrier to nosocomial contamina-
tion (Table IV). Negative-pressure wound therapy and bead
pouch techniques were not employed in this cohort.
Future work in this area should consist of prospective
studiespreferably large, randomized trials. Sample-size calcu-
lations for potential randomized trials must be based on previous
ndings so that potential effects can be estimated, and the present
study provides some such ndings. Additional work is required
in order to identify the particular injury factors that allow for safe
primary closure.
Surgeon judgment regarding the adequacy of debride-
ment and wound closure is still paramount in the treatment of
open fractures; however, primary closure may be preferable for
carefully selected low-grade injuries. Patients must be monitored
closely, regardless of closure type, in order to assess for a surgical-
site infection and to institute timely treatment.
Appendix
A gure comparing the infection rates between the matched
treatment groups is available with the online version of
this article as a data supplement at jbjs.org. n
Richard J. Jenkinson, MD, MSc, FRCS(C)
Alexander Kiss, PhD
Samuel Johnson, MD
David J.G. Stephen, MD, FRCS(C)
Hans J. Kreder, MD, MPH, FRCS(C)
Sunnybrook Health Sciences Centre,
University of Toronto, 2075 Bayview Avenue,
Suite MG-321, Toronto, ON M4N 3M5, Canada.
E-mail address for R.J. Jenkinson: richard.jenkinson@sunnybrook.ca
385
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES
References
1. Gustilo RB, Anderson JT. Prevention of infection in the treatment of one thousand
and twenty-ve open fractures of long bones: retrospective and prospective analy-
ses. J Bone Joint Surg Am. 1976 Jun;58(4):453-8.
2. Hampton OP Jr. Basic principles in management of open fractures. J Am Med
Assoc. 1955 Oct 1;159(5):417-9.
3. Trueta J. Closed treatment of war fractures. Lancet. 1939;233(6043):1452-5.
4. Trueta J. Reections on the past and present treatment of war wounds and
fractures. Mil Med. 1976 Apr;141(4):255-8.
5. Okike K, Bhattacharyya T. Trends in the management of open fractures. A critical
analysis. J Bone Joint Surg Am. 2006 Dec;88(12):2739-48.
6. Benson DR, Riggins RS, Lawrence RM, Hoeprich PD, Huston AC, Harrison JA.
Treatment of open fractures: a prospective study. J Trauma. 1983 Jan;23(1):25-30.
7. Cullen MC, Roy DR, Crawford AH, Assenmacher J, Levy MS, Wen D. Open fracture
of the tibia in children. J Bone Joint Surg Am. 1996 Jul;78(7):1039-47.
8. DeLong WG Jr, Born CT, Wei SY, Petrik ME, Ponzio R, Schwab CW. Aggressive
treatment of 119 open fracture wounds. J Trauma. 1999 Jun;46(6):1049-54.
9. Carsenti-Etesse H, Doyon F, Desplaces N, Gagey O, Tancr` ede C, Pradier C,
Dunais B, Dellamonica P. Epidemiology of bacterial infection during management of
open leg fractures. Eur J Clin Microbiol Infect Dis. 1999 May;18(5):315-23.
10. Gustilo RB, Mendoza RM, Williams DN. Problems in the management of type III
(severe) open fractures: a new classication of type III open fractures. J Trauma.
1984 Aug;24(8):742-6.
11. Salas M, Hofman A, Stricker BH. Confounding by indication: an example of
variation in the use of epidemiologic terminology. Am J Epidemiol. 1999 Jun
1;149(11):981-3.
12. DAgostino RB Jr. Propensity score methods for bias reduction in the
comparison of a treatment to a non-randomized control group. Stat Med. 1998
Oct 15;17(19):2265-81.
13. Austin PC. Primer on statistical interpretation or methods report card on
propensity-score matching in the cardiology literature from 2004 to 2006: a
systematic review. Circ Cardiovasc Qual Outcomes. 2008 Sep;1(1):62-7.
14. Peduzzi PN, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study
of the number of events per variable in logistic regression analysis. J Clin Epidemiol.
1996 Dec;49(12):1373-9.
15. Parsons LS. Performing a 1:N Case-Control Match on Propensity Score. In:
Proceedings of the 29th SAS Users Group International 2004; 2004 May 9-12;
Montreal, Canada. Paper no. 16529. http://www2.sas.com/proceedings/
sugi29/165-29.pdf.
16. Brumback RJ, Jones AL. Interobserver agreement in the classication
of open fractures of the tibia. The results of a survey of two hundred and
forty-ve orthopaedic surgeons. J Bone Joint Surg Am. 1994 Aug;76(8):
1162-6.
17. Orthopaedic Trauma Association: Open Fracture Study Group. A new classi-
cation scheme for open fractures. J Orthop Trauma. 2010 Aug;24(8):457-64.
386
THE J OURNAL OF BONE & J OI NT SURGERY
d
J BJ S . ORG
VOLUME 96-A
d
NUMBER 5
d
MARCH 5, 2014
DELAYED WOUND CLOS URE I NCREAS ES DEEP-I NFECTI ON RATE
AS S OCI ATED WI TH LOWER-GRADE OPEN FRACTURES