1. Wound Repair and Regeneration
Diagnostic validity of three swab techniques for identifying
chronic wound infection
Sue E. Gardner, PhD, RN1,2; Rita A. Frantz, PhD, RN2; Charles L. Saltzman, MD3; Stephen L. Hillis, PhD1;
Heeok Park, MSN, RN2; Melody Scherubel, BSN4
1. Center for Research in the Implementation of Innovative Strategies in Practice (CRIISP), Iowa City VA Medical Center,
2. The University of Iowa College of Nursing,
3. The University of Iowa College of Medicine, and
4. Iowa City VA Medical Center, Iowa City, Iowa
Reprint requests: ABSTRACT
Sue E. Gardner, PhD, RN, 320 NB, The
University of Iowa College of Nursing, Iowa This study examined the diagnostic validity of three different swab techniques in
City, IA 52241-1121. identifying chronic wound infection. Concurrent swab specimens of chronic
Fax: (319) 335-9990; wounds were obtained using wound exudate, the Z-technique, and the Levine
Email: sue-gardner@uiowa.edu technique, along with a specimen of viable wound tissue. Swab and tissue spec-
imens were cultured using quantitative and qualitative laboratory procedures.
Manuscript received: September 26, 2005 Infected wounds were defined as those containing 1Â106 or more organisms per
Accepted in final form: June 1, 2006 gram of tissue. Accuracy was determined by associating the quantitative cultures
of swab specimens with the cultures from tissue specimens using receiver operat-
DOI:10.1111/j.1743-6109.2006.00162.x ing characteristic curves. Of the 83 study wounds, 30 (36%) were infected. Accu-
racy was the highest for swab specimens obtained using Levine’s technique at
0.80. Based on Levine’s technique, a critical threshold of 37,000 organisms per
swab provided a sensitivity of 90% and a specificity of 57%. The mean concord-
ance between swab specimens obtained using Levine’s technique and tissue spec-
imens was 78%. The findings suggest that swab specimens collected using
Levine’s technique provide a reasonably accurate measure of wound bioburden,
given that they are more widely applicable than tissue cultures. The diagnostic
validity of Levine’s technique needs further study using an alternative reference
standard, such as the development of infection-related complications.
Although swab specimens are commonly collected to ex- bioburden of chronic wounds. However, the ability to
amine the bioburden of chronic wounds, the usefulness of draw definitive conclusions from these studies has been
the information provided by cultures based on these spec- problematic due to design and methodological issues.
imens is unclear. Furthermore, health care providers who First, the procedures used to culture, isolate, and identify
collect swab specimens are provided conflicting informa- organisms varied greatly from study to study. Some stud-
tion regarding which collection technique represents ‘‘best ies1 failed to report laboratory procedures, which compro-
practice.’’ The purpose of this study was to examine the mised critical evaluation and sound comparison. Second,
diagnostic validity of three different swab techniques in some of the studies had very small samples1,4 and one used
identifying chronic wound infection. The primary research wound models4 rather than clinical cases. These limita-
questions that were addressed were as follows: What is the tions constrained generalization. Third, most of the studies
accuracy of quantitative cultures based on swab specimens failed to provide a definition of a ‘‘positive’’ culture3,5 or
of wound exudate among a sample of infected and nonin- defined a positive culture as the growth of any organism.
fected, nonarterial, chronic wounds as compared with Therefore, it was impossible to determine whether these
quantitative cultures of viable wound tissue (reference study findings were consistent with our current under-
standard)? What is the accuracy of quantitative cultures standing of wound microbiology: specifically, that all sec-
based on swab specimens obtained using the Z-technique ondary wounds are contaminated with microorganisms,
as compared with the reference standard? What is the ac- although they are not necessarily infected. Comparing
curacy of quantitative cultures based on swab specimens
obtained using Levine’s technique1 as compared with the
reference standard? Are there differences in the accuracy
of quantitative cultures among the three swab techniques? AUC Area under the curve
What is the concordance between qualitative cultures CFU Colony-forming unit
based on swab specimens obtained using wound exudate, HbA1c Hemoglobin A1c
Z-technique, and Levine’s technique and qualitative cul- RBC Red blood cell
tures based on tissue specimens with respect to the simul- ROC Receiver operating characteristic
taneous recovery of organisms? TcPO2 Transcutaneous oxygen
Previous studies1–5 have suggested that swab cultures TSB Tryptic soy broth
may be comparable to tissue cultures in determining the WBC White blood cell
548 Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c
2. Gardner et al. Diagnostic validity of techniques for identifying infection
swab with tissue specimens on criteria that do not corre- presence of a full-thickness, nonarterial chronic wound,
spond with a conceptually sound definition of wound in- white blood cell (WBC) count greater than 1,500 cells/mm3
fection precluded the ability to draw conclusions with or total lymphocyte count greater than 800 cells/mm3,
respect to diagnostic validity. Perhaps the most serious platelet count greater than 125,000/mm, no coagulopa-
methodological problem presented by these studies was thies, and not receiving anticoagulation therapy. The
that the specific techniques used to collect swab specimens wound sample was limited to chronic wounds in order to
were not described.2,4,5 Swabbing techniques vary greatly address the secondary aim of the study, which was to ex-
according to wound preparation (i.e., cleansing or no amine the association of chronic wound infection with risk
cleansing), area of the wound sampled, and duration of factors of infection. The sample was restricted to nonarte-
sampling. Given this variation, the specific technique used rial chronic wounds in order to decrease the risk of infec-
may have been an unrecognized, although significant, con- tion imposed by tissue biopsy in wounds with poor
founding factor on the findings of studies aimed at exam- perfusion. The sample was restricted to full-thickness
ining the diagnostic validity of swab cultures. chronic wounds in order to justify acquisition of full-thick-
This study addressed the limitations of previous studies ness tissue specimens. Patients with low WBC counts were
by delineating and fully describing the swab techniques excluded to reduce the risk of wound infection related to
that were studied and compared, comparing the swab tissue biopsies. Patients with low platelet counts, coagulo-
techniques against the reference standard method of spec- pathies, or anticoagulation therapy were excluded to re-
imen collection (i.e., biopsy of ‘‘viable’’ wound tissue), ap- duce the risk of bleeding associated with tissue biopsy.
plying a standard, research-based definition of positive Eligible subjects were invited to enter the study, and in-
culture, adopting and describing microbiological proce- formed consent was obtained from the subject or legal rep-
dures to enhance the recovery and quantification of or- resentative. One wound was randomly selected when
ganisms consistent with this definition, and using data subjects had more than one eligible wound.
analysis strategies compatible with addressing diagnostic The primary study variables were culture findings based
validity. Although the sample was limited to chronic on swab specimens of wound exudate, swab specimens us-
wounds, the results of this study are applicable to all ing a Z-stroke technique, swab specimens using the Levine
wounds healing by secondary intention, because diagnos- technique, and specimens of viable wound tissue. Swab
tic accuracy should not be influenced by wound etiology. and tissue specimens were processed and cultured at a sin-
A secondary aim of this study was to examine patient gle microbiology laboratory dedicated to microbiological
and wound variables believed to be risk factors of chronic research studies.
wound infection. The association between these factors
and infection status, as measured by quantitative cultures Processing of tissue samples
of wound tissue, will support the generation of hypotheses
for future studies. The laboratory procedure for culturing tissue specimens
was comparable with that suggested by Krizek and Rob-
MATERIALS AND METHODS son.6 Tissue specimens were weighed, homogenized, and
serially diluted in tryptic soy broth (TSB; Remel, Lenexa,
An observational, cross-sectional study design was used KS). Each dilution was plated onto Columbia blood agar
for this investigation. Concurrent swab specimens were (Remel), BBLtCHROM-agart Candida (BD, Sparks,
obtained from a sample of chronic wounds using wound MD), eosin-methylene blue agar (Remel), and reduced
exudate, Z-technique, and Levine’s technique. The diag- agar media. Reduced agar plates were placed in an anaer-
nostic validity of each technique was determined by asso- obic chamber and incubated in ambient air at 37 1C for 48
ciating the quantitative culture findings of swab specimens hours. CHROM-agart Candida plates were placed in an
obtained with each swab technique with quantitative cul- aerobic chamber and incubated at 30 1C for optimal yeast
ture findings of concurrently collected specimens of viable growth. All other plates were incubated under aerobic
wound tissue. In addition, the concordance of each swab conditions in 5% CO2 at 37 1C for 48 hours for optimal
technique was described by comparing the qualitative cul- growth of Streptococcus species and other fastidious or-
ture findings of swab specimens with the qualitative cul- ganisms. To provide qualitative culture data, organisms
ture findings of tissue specimens in order to describe more were identified using standard microbiological procedures
completely the performance of each swab technique. The based on criteria such as colony morphology and gram
sequence and timing of specimen acquisition was designed stain appearance.7 For example, Staphylococcus aureus
to minimize alterations in culture findings related to pre- was identified based on characteristic yellow B-hemolytic
ceding manipulations of the wound bed or the progression colonies on Columbia blood agar, which on stain appeared
of time. as gram-positive cocci organized into grape-like clusters
and that tested catalase positive and staph latex positive.
Setting, sample collection, and study variables
Streptococci were identified to Lancefield groups (A, B, C,
G, and F) by agglutination with appropriate antisera using
The study population consisted of patients with nonarte- the PathoDx kit (Remel).
rial chronic wounds. A Department of Veteran’s Affairs To provide quantitative culture data, each identified
Medical Center and a university-associated tertiary hospi- organism was quantified by counting the number of colo-
tal served as settings for the study. The institutional review ny-forming units (CFU) on each plate. Because tissue
board approved the study protocol before enrolling sub- specimens were based on weight of tissue, the plate count
jects. Subjects were screened and enrolled in the study multiplied by the dilution factor yielded the number of or-
based on the following criteria: 18 years of age or older, ganisms per gram of tissue.
Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c 549
3. Diagnostic validity of techniques for identifying infection Gardner et al.
The ‘‘true’’ infection status of the wound was based on their association with chronic wound infection. These var-
the quantitative culture findings from tissue specimens. In- iables include age, gender, race, diabetes, systemic antibi-
fected wounds were defined as those with 1Â106 or more otic therapy, red blood cell count (RBC), WBC, albumin
organisms per gram of viable tissue.8 Similarly, noninfect- level, hemoglobin A1c (HbA1c) levels for subjects with di-
ed wounds were defined as those with less than 1Â106 or- abetes, type of chronic wound, size of the wound, depth of
ganisms per gram of viable wound tissue. Although the the wound, duration of study ulcer, amount of necrotic
term ‘‘greater than 105’’ has been interpreted to mean both tissue, wound tissue oxygen (transcutaneous oxygen
1Â105 9 and 1Â106 8 organisms per gram of tissue, 1Â106 [TcPO2] measurement), and type of wound dressing.
was the intended critical value (M. C. Robson, personal Demographic, medical history, and medication data (in-
communication, May 2002). cluding systemic antibiotics) were collected from the pa-
Unlike quantitative culture findings, little or no evi- tient record and patient/caregiver report. Data regarding
dence exists from which to define ‘‘true’’ infection status the history, type, and location of the study wound as well
from qualitative culture findings. That is, it is unclear as the type of wound dressing were recorded. Blood sam-
which organisms represent a definitive threat to the wound ples were sent for laboratory analyses of HbA1c levels,
and, therefore, which constitute infection when cultured complete blood count, and albumin levels.
from a wound.10 Given that chronic wounds are often po- Wound tissue oxygenation was measured using a TcPO2
lymicrobial with many organisms considered surface con- monitor (Novametric Model 840 Novametrix Systems
taminants only, it is difficult to discern which species must Inc., Wallingford, CT). These monitors measure arterial
be recovered and identified in wound specimens. Of the oxygen partial pressure noninvasively in the tissue adja-
numerous organisms that colonize chronic wounds, ex- cent to the wound. The TcPO2 sensor was secured on the
perts believe S. aureus, Pseudomonas aeruginosa, Groups skin just proximal to the study wound and equilibrated for
A and B Streptococci, and anaerobes to be the primary 20 minutes before recording TcPO2 values. The TcPO2 lev-
cause of delayed healing and infection,10 but there is little els displayed on the monitor panel were recorded after
evidence to support this assertion. It is also unclear wheth- equilibration. TcPO2 values were recorded at 1-minute in-
er or not the number of different species colonizing a tervals over a 5-minute period.
wound is indicative of infection status. Some evidence sug- After removing the dressing, wound exudate apparent
gests that chronic wounds with multiple species of organ- on the wound surface was sampled with a Culturettes
isms have poor healing outcomes.11 Despite the inability to swab (BD, Cockeysville, MD; Specimen 1: wound exudate)
use qualitative culture findings from tissue specimens to and placed in a labeled transport container. The wound
define ‘‘true’’ wound infection status, qualitative culture was then cleansed by gently rubbing the surface with
findings from tissue specimens can be used as the reference nonbacteriostatic saline. Following cleansing, a second
standard for examining the performance of swab speci- Culturettes swab was applied to the wound surface in a
mens obtained from each swab technique because speci- close zig–zag manner covering the entire injured area while
mens of viable wound tissue are considered the most valid simultaneously rotating the swab between the thumb and
wound specimen.12 forefinger (Specimen 2: Z-technique). The swab was then
placed in a labeled transport container. Next, an area near
Processing swab specimens the center of the wound free of necrotic tissue and debris
was cleansed with nonbacteriostatic saline. The end of a
Procedures for laboratory processing of swab specimens third Culturettes swab was rotated over a 1 cm2 area for
consisted of placing them in 1 mL of sterile saline and 5 seconds with sufficient pressure to extract fluid from
vortexing them for 15 seconds. They were then serially di- within the wound tissue (Specimen 3: Levine’s technique)
luted in TSB (Remel) and each dilution was plated onto and placed in a labeled transport container.
Columbia blood agar (Remel), BBLtCHROM-agart Wound tissue biopsy (Specimen 4: wound tissue) was
Candida (BD), and eosin–methylene blue agar (Remel). then performed according to the procedures outlined by
CHROM-agart Candida plates were placed in an aerobic Stotts.13 After cleansing the wound, a specimen of viable
chamber and incubated at 30 1C for optimal yeast growth. wound tissue (approximately 1 g) was removed from under
All other plates were incubated under aerobic conditions the area of the wound sampled with swab #3 using a 4–
in 5% CO2 at 37 1C for 48 hours for optimal growth of 6 mm dermal punch instrument. A sterile technique was
Streptococcus species and other fastidious organisms. used during the procedure. The specimen was placed in a
Swab specimens were not incubated under anaerobic con- labeled transport container.
ditions because swab specimens are believed to be unsuit- Swab and tissue specimens were transported to the
able for anaerobic culture.7 To provide qualitative culture microbiology laboratory for quantitative and qualitative
data, isolated organisms were identified as described cultures. Laboratory technicians were kept blind to the
above. The quantification of each organism was accom- study aims and study procedures in order to minimize ob-
plished by counting the number of CFU on each plate. server bias. All specimens were processed within 2 hours of
Because dilutions were based on one swab, the plate count acquisition in order to minimize alterations in number of
times the dilution factor yielded the total number of organisms related to progression of time. Culture media
organisms per swab. and isolation criteria were standardized for all specimens
in order to decrease bias associated with differences in lab-
Data collection oratory methodology.
The wound margin outline was then traced on a trans-
Secondary study variables (risk factors) were also meas- parent film with an indelible marker. Tracings were labeled
ured as a part of study procedures in order to examine with subject identification number and date. The depth of
550 Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c
4. Gardner et al. Diagnostic validity of techniques for identifying infection
the wound was measured using a cotton-tipped swab informative test will have an ROC curve that lies along the
placed in the deepest portion of the wound. The point on diagonal and hence have an AUC of approximately 0.5.
the swab that was level with the peri-wound skin was A very informative test will have an ROC curve close to
marked and measured with a centimeter ruler. The amount the upper left corner and an AUC close to 1. Ninety-five
of necrotic tissue in the wound bed was rated using direct percent confidence intervals (95% CI) were computed for
observation and quantified using a Likert-type scale that the AUC of each swab technique and the null hypothesis
classified percentage of the wound bed covered with ne- (H0: AUC50.5) was statistically tested (a level 0.05; two
crotic tissue.14 A photographic image of the wound was tailed) using the algorithm described by Delong, Delong,
taken with a digital camera. and Clarke-Pearson.16 Computations were performed us-
ing Statas 8.1 for Windowss (StataCorp, College Station,
TX).
Statistical analysis
Subject and wound data were entered into electronic dat- Comparison of accuracy among the swab techniques
abases, with the data cleaned using range and consistency To address research question 4, we tested the global null
checks. Each wound was classified as infected or nonin- hypothesis of equality of the three AUCs, followed by
fected based on quantitative culture findings from wound pairwise comparisons if the global null hypothesis was re-
tissue specimens using the definition described above. This jected. These tests were performed using the algorithm de-
classification represented the reference standard criterion scribed by Delong et al.16 All tests were two tailed with
from which to compare the performance of quantitative a50.05.
cultures of swab specimens obtained using each swab To assess the practical usefulness of the swab tech-
technique. Data on the type and number of organisms niques, estimated specificities were tabulated correspond-
per swab were also entered for each of the three swab ing to a range of sensitivity levels (i.e., 0.90, 0.80, 0.70), as
techniques. well as the corresponding estimated thresholds and posi-
tive and negative predictive values. For a given sensitivity,
the specificity was estimated by linear interpolation from
Summary statistics the fitted ROC curve and the corresponding threshold was
Subject and wound databases were used to compute sum- estimated by linear interpolation of the thresholds corre-
mary statistics of the subject and wound sample and to sponding to the nearest lower and higher observed specifi-
compute inferential statistics for examining differences be- cities. Corresponding positive predictive values (PPV) and
tween the infected and noninfected groups. Nominal var- negative predictive values (NPV) were estimated from the
iables were statistically described with percentages and sensitivity, estimated specificity, and sample prevalence
compared between the infected and noninfected groups rate of infected wounds in the sample using the following
using a chi-square test for independence or Fisher’s exact relationships:17
test. Continuous variables were described with means and
standard deviations, and statistically compared between
the infected and noninfected groups using the nonpara- PPV ¼ p sensitivity=
metric Wilcoxon rank-sum test. Medians were also de- ½p sensitivity þ ð1 À pÞð1 À specificityÞŠ
scribed for continuous variables having outliers. Given the
exploratory nature of these statistical inference tests, no NPV ¼ ð1 À pÞ specificity=
correction was made for multiple comparisons. A two- ½ð1 À pÞspecificity þ pð1 À sensitivityÞŠ
tailed a level of 0.05 was used to identify significant differ-
ences between the infected and noninfected groups. where p is the sample prevalence rate. The threshold of 1Â
106 organisms per swab was also specifically examined in
Accuracy of the three swab techniques order to address specifically the most common threshold
To address research questions 1–3, receiver operating reported in the literature.
characteristic (ROC) curves, computed using the nonpar-
ametric trapezoidal approach,15 were used to evaluate the Concordance of the swab techniques
diagnostic validity of each swab technique. This approach To address research question 5, qualitative cultures from
is well suited in this case because the number of organisms tissue specimens were compared with the qualitative cul-
per swab is a continuous rather than discrete variable. An tures from the swab specimens of each swab technique.
ROC curve is a plot of sensitivity vs. specificity for each First, qualitative culture findings were described for tissue
value for the predictive variable and is hence a graphical specimens and each swab technique. The mean number of
way of displaying the complete performance of the swab different species isolated per wound was calculated as well
techniques. as the frequency with which the organisms, S. aureus,
The area under the ROC curve (AUC) provides a sum- P. aeruginosa, and Groups A and B streptococci were
mary measure of accuracy for each swab technique: it es- recovered. The frequency with which anaerobes were
timates the accuracy rate for discriminating between recovered from tissue specimens was also tabulated. The
infected and noninfected wounds.15 It estimates the prob- recovery of anaerobes from swab specimens could not be
ability that a randomly selected patient with an infected tabulated because swab specimens were not incubated
wound (according to the reference standard) will have a under anaerobic conditions.
higher outcome (number of organisms per swab) than a Second, concordance between the qualitative cultures
randomly selected patient with a noninfected wound. The from tissue and swab specimens was examined. For each
AUC for an ROC curve is always between 0 and 1. A non- study wound, the concordance in recovering all organisms
Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c 551
5. Diagnostic validity of techniques for identifying infection Gardner et al.
was determined using the following equation2: the number perficial or small to justify a full-thickness punch biopsy.
of specific isolates simultaneously recovered from the tis- Another 57 patients (27% of screened patients) were ex-
sue and the swab specimen divided by the total number of cluded because patient conditions placed them at risk for
isolates collected by either specimen, multiplied by 100. biopsy-related complications such as bleeding.
The mean concordance was then calculated for each swab
technique. In addition, the concordance between tissue
and swab specimens in specifically recovering S. aureus, P. Demographic and wound characteristics
aeruginosa, and Groups A and B streptococci was comput- Eighty-three subjects participated in the study. Sixty-eight
ed as total (i.e., total number of concordant observations/ (82%) of the subjects were male, and 81 (98%) were white.
total number of paired observations), occurrence (i.e., Descriptive statistics of subject characteristics are present-
number of positive concordant observations/total number ed in Table 1 for the total, infected, and noninfected
of wounds in which the specific organism was recovered groups. Comparison of the infected and noninfected
from either or the tissue or swab specimen), and nonoc- groups revealed no significant differences in age, gender,
currence (i.e., number of negative concordant observa- race, diabetes status, RBC, WBC, and albumin level be-
tions/total number of wounds in which the specific tween the groups. Treatment with systemic antibiotic
organism was absent from either the tissue or swab spec- treatment was associated with having a noninfected
imen) concordance. Three measures of concordance were wound (Fisher’s exact p50.011). Among the 73 subjects
calculated because estimates are influenced by the manner (88%) with Type 1 or Type 2 diabetes, there was no signif-
in which concordance is defined and total concordance can icant difference in HbA1c levels between the infected and
be inflated by a high percentage of nonoccurrence agree- noninfected groups.
ments when few occurrence agreements occur. Wound characteristics for the total group, infected, and
noninfected group are summarized in Table 2. Of the 83
RESULTS wounds, 30 (36%) were infected based on quantitative tis-
sue cultures (specimen 4). Diabetic foot ulcers were the
Of 310 patients who were screened for the study, 102 most common type of wound. Comparisons of the infected
(33%) were eligible based on exclusion criteria. Of the 102 and noninfected groups showed no significant difference
eligible subjects, 83 (81%) consented to study enrollment in wound type, wound depth, wound duration, amount of
and finished study participation. Patients who declined to necrotic tissue, or type of wound dressing between the two
enroll often did so because they objected to having tissue groups.
removed from their wound for tissue cultures (n519; 19% There was a significant difference in wound size between
of eligible subjects). Of the 208 patients who were excluded the infected and noninfected groups (rank sum p50.009),
from study participation, most (n5136; 65% of screened with the noninfected group having the larger surface area
patients) were excluded because their wounds were too su- (cm2). Further analyses revealed that subjects being
Table 1. Patient characteristics for the total, infected, and noninfected groups
Characteristic Total sample Infected group Noninfected group p-value
n 83 30 53
Mean age Æ SD (years) 57 Æ 11.9 55 Æ 11.8 58 Æ 12.0 0.257
Gender 1.000
Male 68 (82%) 25 (83%) 43 (81%)
Female 15 (18%) 5 (17%) 10 (19%)
Race 0.533
White 81 (98%) 30 (100%) 51 (96%)
Black 2 (2%) 0 (0%) 2 (4%)
Diabetes 0.735
None 10 (12%) 4 (13%) 6 (11%)
Type 1 11 (13%) 5 (17%) 6 (11%)
Type 2 62 (75%) 21 (70.0%) 41 (78%)
Systemic antibiotics 0.011n
No 48 (58%) 23 (77%) 25 (47%)
Yes 35 (42%) 7 (23%) 28 (53%)
Mean RBC Æ SD (million cells/mL) 4.2 Æ 0.60 4.3 Æ 0.59 4.2 Æ 0.61 0.388
Mean WBC Æ SD (cells/mL) 8586.7 Æ 2711.03 8610 Æ 2116.33 8573.6 Æ 3015.23 0.566
Mean albumin Æ SD (g/dL) 3.7 Æ 0.51 3.8 Æ 0.48 3.7 Æ 0.52 0.494
Mean HbA1c Æ SD 7.9 Æ 1.67 (n573) 8.0 Æ 1.76 (n526) 7.9 Æ 1.63 (n547) 0.632
n
p 0.05.
552 Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c
6. Gardner et al. Diagnostic validity of techniques for identifying infection
Table 2. Wound characteristics of the total, infected, and noninfected groups
Total sample Infected group Noninfected group p-value
n 83 30 53
Wound type 0.378
Diabetic ulcer 64 (77.1%) 25 (83.3%) 39 (73.6%)
Pressure ulcer 6 (7.2%) 0 (0%) 6 (11.3%)
Venous ulcer 5 (6.0%) 2 (6.7%) 3 (5.7%)
Secondary incision 4 (4.8%) 1 (3.3%) 3 (5.7%)
Trauma 4 (4.8%) 2 (6.7%) 2 (3.8%)
Mean size Æ SD (cm2) 11.3 Æ 22.41 5.4 Æ 10.02 14.6 Æ 26.55 0.009n
Median size (cm2) 3.6 1.88 5.1
Mean depth Æ SD (cm) 0.7 Æ 0.89 0.6 Æ 0.51 0.8 Æ 1.05 0.909
Median depth (cm) 0.4 0.4 0.5
Mean duration of wound Æ SD (weeks) 82.6 Æ 328.74 97.8 Æ 281.28 74.0 Æ 355.08 0.064
Median duration 15.0 24.0 13.0
Amount of wound bed with necrotic tissue 0.543
None 51 (62%) 20 (67%) 31 (59%)
1–50% 26 (31%) 9 (30%) 17 (32%)
50–100% 6 (7%) 1 (3%) 5 (9%)
Mean wound tissue oxygen Æ SD (mmHg) 44.6 Æ 15.90 46.5 Æ 15.82 43.4 Æ 16.00 0.862
Primary wound dressing 0.145
None 2 (2%) 2 (7%) 0 (0%)
Dry 23 (28%) 7 (23%) 16 (30%)
Moisture retentive 58 (70%) 21 (70%) 37 (70%)
n
p 0.05.
treated with systemic antibiotics had wounds with larger tistically significant in terms of their agreement with the
surface areas. The mean size of the wounds being treated reference standard criterion: quantitative cultures of
with systemic antibiotics was 14.5 cm2 (SD Æ 22.99), while wound tissue.
the mean size of the wounds not being treated with sys-
temic antibiotics was 8.8 cm2 (SD Æ 21.87); the respective
medians were 7.9 and 2.00 cm2. The difference in wound
size between those being treated with systemic antibiotics
and those not being treated was statistically significant
(rank sum p50.014), suggesting that antibiotics may ex-
plain the association between wound size and infection
status.
However, in a rank-sum stratified analysis (using the
Cohran–Mantel–Haenszel statistic), wound size remained
significantly higher in noninfected patients (CMH p5
0.0255) after controlling for systemic antibiotics.
Accuracy of the three swab techniques
The ROC curves for each of the three swab techniques are
plotted in Figure 1 along with the diagonal reference. As
shown in the figure, Levine’s technique (specimen 3) has
the largest AUC. In addition, both the wound exudate
(specimen 1) and the Z-technique (specimen 2) are above
the diagonal reference line.
Table 3 presents the AUC for the three swab techniques
and the results of statistical tests to examine whether the Figure 1. ROC curves for three swab techniques. Wound exu-
AUC for each swab technique is significantly greater than dates, specimen 1; Z-technique, specimen 2; Levine’s tech-
0.5 (i.e., diagonal reference). All swab techniques were sta- nique, specimen 3.
Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c 553
7. Diagnostic validity of techniques for identifying infection Gardner et al.
Table 3. AUC for three quantitative swab techniques
95% Confidence interval
Swab technique AUC (Standard error) Significance levelz Lower bound Upper bound
Wound exudates 0.67 ( Æ 0.060) 0.011 0.55 0.79
Z-technique 0.67 ( Æ 0.063) 0.010 0.55 0.80
Levine’s technique 0.80 ( Æ 0.052) 0.001 0.70 0.90
z
Null hypothesis: true area 5 0.5.
Comparisons of accuracy among the swab techniques sensitivities and specificities corresponding to the critical
threshold of 1Â106 organisms per swab.
As shown in Table 3, the AUC for Levine’s technique
(specimen 3) is much larger than the AUC for the other Concordance of the three quantitative swab techniques
two swab techniques. The test of the null hypothesis of
equality of the three AUCs was significant (w256.15, 2 Qualitative culture findings from the tissue and swab spec-
p50.0462), implying that the three population AUCs are imens are presented in Table 5. The mean number of dif-
not all equal. Follow-up tests showed a statistically signif- ferent species recovered per wound was similar across all
icant difference between Levine’s technique and the Z- specimens and ranged from 3.0 to 3.5 species per wound.
technique (z52.13, p50.0326) and a difference between S. aureus was recovered more frequently than either P. ae-
Levine’s technique and wound exudate that approached ruginosa or Groups A and B Streptococci in this sample of
statistical significance (z51.94, p50.0525). wounds. Further, the proportion of wounds from which
Table 4 presents a range of sensitivities (i.e., 0.90, 0.80, these specific organisms were recovered was also similar
and 0.70) for the three swab techniques along with the cor- across all specimens. Anaerobes were recovered in only
responding interpolated specificities, critical thresholds, five (6%) study wounds.
and positive and negative predictive values. As expected Concordance analyses of qualitative cultures are pre-
from the display of the ROC curves and the ROC AUC sented in Table 6. Concordance values were high for all
analyses above, Levine’s technique provides the highest swab techniques with respect to recovering all organisms.
specificity for a given sensitivity. For example, Levine’s Each swab technique was highly concordant in recovering
technique, using a critical threshold of 3.7Â104 organisms S. aureus as indicated by all measures of concordance (i.e.,
per swab, had a sensitivity of 90% and a specificity of total, occurrence, and nonoccurrence). Occurrence con-
57%; in comparison, the wound exudate and Z-technique cordance in recovering Groups A and B Streptococci was
have specificities of 40% and 27%, respectively, corre- the lowest concordance measure for all swab techniques at
sponding to a sensitivity of 90%. Table 4 also presents the 67%. In general, concordance values for Levine’s
Table 4. Range of sensitivities with corresponding indices for each swab technique
z §
Swab technique Sensitivity Specificity Critical threshold PPV NPV
Wound exudate 0.70 0.47 75,000 0.43 0.74
0.80 0.40 23,666 0.43 0.78
0.90 0.40 20,000 0.46 0.88
0.43 0.75 1,000,000 0.50 0.70
Z-technique 0.70 0.47 600,000 0.43 0.74
0.80 0.36 300,000 0.41 0.76
0.90 0.27 62,500 0.41 0.83
0.63 0.53 1,000,000 0.43 0.72
Levine’s technique 0.70 0.72 500,000 0.58 0.81
0.80 0.57 100,000 0.51 0.83
0.90 0.57 37,000 0.54 0.91
0.57 0.91 1,000,000 0.77 0.79
z
PPV, positive predictive value.
§
NPV, negative predictive value.
For each swab technique, the first three rows give the interpolated specificity, critical threshold, PPV, and NPV corresponding to the
specified sensitivity (i.e., 0.70, 0.80, or 0.90). The fourth row gives the sensitivity, specificity, PPV, and NPV corresponding to a
critical threshold of 1,000,000 organisms per swab.
554 Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c
8. Gardner et al. Diagnostic validity of techniques for identifying infection
Table 5. Qualitative culture findings by specimen type
Wound exudate Z-technique Levine’s Tissue
Mean (SD) number of species/wound 3.0( Æ 1.95) 3.5( Æ 1.98) 3.0( Æ 1.93) 3.1( Æ 1.89)
Number (%) wounds with Staphylococcus aureus 44 (53%) 44 (53%) 45 (54%) 46 (55%)
Number (%) wounds with Pseudomonas aeruginosa 12 (15%) 15 (18%) 11 (13%) 10 (12%)
Number (%) wounds with Group A or B Streptococci 2 (2%) 2 (2%) 2 (2%) 3 (4%)
Number (%) wounds with anaerobes n
NA NA NA 5 (6%)
n
NA, not available.
N583.
technique were equal to or higher than both the wound Systemic antibiotic use was the only subject variable
exudate and the Z-technique in all instances except mean found to be associated with wound infection status, and
concordance. this finding is also consistent with the findings of our pre-
vious study.19 Subjects treated with systemic antibiotics
were less likely to have an infected wound than those who
were not being treated with systemic antibiotics. Subjects
DISCUSSION in this sample were being treated with systemic antibiotics
Based on quantitative wound tissue cultures, 36% of for cellulitis, osteomyelitis, and bacteremia. Systemic anti-
wounds in the sample were infected. It is important to biotics appear to be relatively effective in nonarterial
note that, for this study, infection was specifically defined chronic wounds because only 20% of wounds treated with
as 1Â106 organisms or more per gram of tissue in order to systemic antibiotics had 1Â106 or more organisms per
avoid the confusion that is associated with the term ‘‘great- gram of tissue when cultured.
er than 105.’’ In their study of chronic wounds, Bill et al.18 Wound area was the only wound variable associated
reported an infection rate of 74% using ‘‘greater than 105’’ with infection status. Wounds with a larger surface area
as their definition of infection. It is unclear whether they were less likely to be infected. This finding was not ex-
used a critical value of 1Â105 or 1Â106 organisms per plained by the fact that larger wounds are more likely to
gram of tissue. The infection rate found in this study be treated with systemic antibiotics. Unlike the findings
(36%) is consistent with the 31% infection rate found in from our previous study,19 wound TcPO2 and amount of
our previous study.19 necrotic tissue were not associated with wound infection
status.
Examination of the recruitment data substantiated the
Table 6. Concordance of qualitative swab and tissue cultures potential value of a valid, noninvasive method of assess-
ing wound bioburden. Analyses of recruitment data
Wound showed that 100% of the eligible patients who refused to
exudate Z-technique Levine’s participate in this study did so because they objected to
(%) (%) (%) wound tissue biopsy. In addition, 93% of screened pa-
tients were excluded from eligibility because they had
All organisms conditions that precluded justification of the risks associ-
Mean concordance 83 78 78 ated with wound biopsy such as bleeding. Although these
Staphyloccus aureus barriers to obtaining tissue cultures are well known, the
accuracy of swab specimens remained unclear despite the
Total concordance 95 93 96
findings of previous studies designed to examine their
Occurrence concordance 91 88 94 accuracy.
Nonoccurrence 90 85 92 The findings of this study indicate that swab specimens
concordance collected from wounds using Levine’s technique per-
Pseudomonas aeruginosa formed better than swab specimens collected using either
the wound exudate or Z-technique. Equally important, the
Total concordance 95 92 96
findings suggest that swab specimens obtained using
Occurrence concordance 69 56 75 Levine’s technique and processed using quantitative labo-
Nonoccurrence 95 91 96 ratory procedures are acceptably accurate when compared
concordance with the quantitative cultures of wound tissue.
Groups A or B Streptococcus This is the first study to examine explicitly different
thresholds for quantitative swab cultures and their corre-
Total concordance 99 99 99
sponding sensitivity and specificity values. For Levine’s
Occurrence concordance 67 67 67 technique, a critical threshold of 3.7Â104 organisms per
Nonoccurrence 99 99 99 swab provided a sensitivity of 90% and a specificity of
concordance 57%. In their study, which compared quantitative swab
cultures obtained with Levine’s technique with quantita-
N583. tive tissue cultures, Bill et al.18 reported data that can be
Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c 555
9. Diagnostic validity of techniques for identifying infection Gardner et al.
interpreted as having a sensitivity of 79% and a specificity genicity of organisms) are important in defining wound
of 60% using the critical threshold of ‘‘greater than 105’’ infection. Microbial load refers to the number of organ-
among a sample of 38 wounds. Unfortunately, Bill et al. isms per gram of tissue or swab, microbial diversity refers
did not examine the sensitivity and specificity for different to the number of different species present in the wound,
threshold values, nor did they compute the accuracy rate and pathogenicity of organisms refers to the type of or-
for discriminating between infected and noninfected ganisms present, for example S. aureus, P. aeruginosa,
wounds. Furthermore, it is also unclear which interpreta- Groups A and B Streptococci, or anaerobes. This study
tion of ‘‘greater than 105’’ was used in their study: 1Â105 primarily focused on microbial load (i.e., number of or-
or 1Â106 organisms per swab. ganisms per gram of tissue), which can only be measured
Although this is the first study to examine different with quantitative cultures, because it has been repeatedly
thresholds for quantitative swab cultures, it does not fully shown to be an important dimension of wound biobur-
address the issue of which threshold provides the best bal- den9; however, other dimensions may also be important.
ance between sensitivity and specificity for use in clinical The accuracy with which qualitative Levine’s cultures
practice or whether the sensitivity and specificity values measure the other two dimensions of wound bioburden,
found in this study are adequate. Selecting a threshold for microbial diversity and pathogenicity of organisms, was
clinical practice requires that the value of identifying most addressed by describing their concordance with qualitative
or all infected wounds be balanced against the costs asso- tissue cultures. Although the mean concordance between
ciated with the number of false positives generated from tissue specimens and swab specimens collected from
this value. wound exudate was slightly higher than swab specimens
We believe that a sensitivity value of 90% and a collected using Levine’s technique, the concordance be-
specificity of 57% are acceptable because swab specimens tween Levine’s technique and tissue specimens was still
obtained with Levine’s technique will enable a wider vari- high at 78%. This finding is consistent with the 74.5%
ety of wounds to be monitored for wound bioburden than concordance reported by the Sapico et al.2 in a study that
tissue cultures. In addition, Levine’s technique will be compared swab and tissue cultures. With respect to the or-
much more practical for repeating cultures in suspicious ganisms of S. aureus, P. aeruginosa, and Groups A and B
wounds that produce negative findings initially than tissue Streptococci, swab specimens obtained using Levine’s
cultures. Although a specificity of 57% means that 43% of technique were more concordant with tissue specimens
noninfected wounds will be identified as infected, the abil- than specimens collected by either of the other two swab
ity to identify accurately 57% of all noninfected wounds techniques. Moreover, swab specimens collected using Le-
represents a substantial improvement over current prac- vine’s technique had high total, occurrence, and nonoc-
tice. Many chronic wounds are indiscriminately treated for currence concordance for recovering S. aureus and high
infection because of the commonly held belief that high total and nonoccurrence concordance for recovering P.
wound bioburden is a significant contributor to delayed aeruginosa and Groups A and B Streptococci. The occur-
healing in chronic wounds. Because evidence was lacking rence concordance for recovering P. aeruginosa and
to support the use of more practical swab cultures, from Groups A and B Streptococci was not as high, but may be
which high bioburden could be verified or ruled out as the due to the low frequency with which these organisms were
contributor to poor healing, many advocated12 that chron- recovered in the wound sample. S. aureus, on the other
ic wounds failing to progress towards closure should be hand, was recovered in over 50% of the wounds. Unfortu-
treated for high wound bioburden.10,20 Swab cultures ob- nately, we could not examine the performance of Levine’s
tained with Levine’s technique would significantly im- technique in recovering anaerobes because standard swab
prove the ability to verify which of these stalled wounds transport procedures are not conducive to the recovery of
have high bioburden, thus decreasing the number receiving these organisms. However, the recovery of anaerobes from
unnecessary treatment. Although we would prefer that the swab specimens has been described through the use of an
specificity be higher, this may be an acceptable trade-off at anaerobic transport container21 and would allow exami-
this point in time, given that it represents an improvement nation of the concordance between tissue and swab spec-
in our knowledge of wound bioburden measurement, and imens with respect to anaerobe recovery. Despite the
because a high sensitivity is most important because inability of this study to examine anaerobe recovery, the
wound infection can lead to serious complications, such findings provide substantial evidence that swab specimens
as amputation. collected using Levine’s technique are concordant with tis-
Although using swab cultures obtained using Levine’s sue specimens and are acceptably accurate in identifying
technique to measure wound bioburden represents an im- microbial diversity and pathogenicity.
provement over current practice, we believe that the po- The limitations related to the reference standard used
tential to improve both the sensitivity and specificity of in this study can be addressed by further examining the
this method is promising. For example, the major limita- diagnostic validity of Levine’s technique using an alter-
tion of this study was that the reference standard used to native reference standard, such as the development of
classify wounds as infected or noninfected (i.e., quantita- infection-related complications. The use of infection-relat-
tive cultures of wound tissue) is imperfect. As noted be- ed complications as the reference standard would allow
fore, the threshold for infection has been widely more precise identification of the optimum critical thresh-
misinterpreted in the literature, and there remain ques- old, and moreover, empirical examination of which wound
tions as to the appropriate definition of wound infection bioburden dimensions (i.e., microbial load, microbial di-
when based solely on wound bioburden measures.10,20 versity, or pathogenicity) are most useful in predicting
These questions include which wound bioburden dimen- wound outcomes and, hence, are more valid definitions of
sions (i.e., microbial load, microbial diversity, or patho- wound infection. Using a swab technique that is a practical,
556 Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c
10. Gardner et al. Diagnostic validity of techniques for identifying infection
yet accurate, method for collecting wound specimens will 8. Heggers JP. Variations of a theme. In: Heggers JP, Robson
facilitate the conduct of studies needed to develop evidence- MC., editors. Quantitative bacteriology: Its role in the arma-
based wound infection definitions that can be applied in a mentarium of the surgeon. Boca Raton: CRC Press, 1991:
wide variety of clinical settings, while concurrently improv- 15–23.
ing the accuracy of the swab technique. 9. Robson MC. Wound infection: a failure of wound healing
Finally, the sensitivity and specificity of swab cultures caused by an imbalance of bacteria. Surg Clin North Am 1997;
obtained with Levine’s technique may also be enhanced 77: 637–50.
through further study of other clinical variables that may 10. Bowler PG, Duerden BI, Armstrong DG. Wound microbiol-
serve as covariates in analyses. Our long-term goal is to ogy and associated approaches to wound management. Clin
develop a multivariable clinical prediction rule based on Microbiol Rev 2001; 14: 244–69.
clinical data, including culture findings, that accurately 11. Trengrove NJ, Stacey MC, McGechie DF, Stingemore NF,
identifies chronic wounds at risk for developing infection- Mata S. Qualitative bacteriology and leg ulcer healing.
related complications, thus leading to prevention of these J Wound Care 1996; 5: 277–80.
undesirable outcomes. 12. Bergstrom N, Allman RM, Alvarez OM, Bennett MA, Carl-
son CE, Frantz RA, Garber SL, Jackson BS, Kaminski MV,
Kemp MG, Kroskop TA, Lewis VL, Makelbust J, Margolis
ACKNOWLEDGMENTS DJ, Marvel EM, Reger SI, Rodeheaver GT, Salcido R,
Xakellis GC, Yarkony GM. Treatment of pressure ulcers.
This study was funded by the Department of Veterans Af-
AHCPR publication No. 95-0652. Rockville, MD: Agency
fairs, Health Services Research and Development, Nursing
for Health Care Policy and Research, Public Health
Research Initiative (NRI-01-005-1), and the National In-
Service, U.S. Department of Health and Human Services,
stitutes of Health, National Institute of Nursing Research
1994.
(NINR RO1 NRO7721). The views expressed in this arti-
13. Stotts NA. How to culture a wound and do a punch biopsy.
cle are those of the authors and do not necessarily repre-
Paper presented at the Clinical Symposium on Wound Man-
sent the views of the Department of Veterans Affairs.
agement, Dallas, TX, 1997.
14. Bates-Jensen BM, Vredevoe D, Brecht ML. Validity and re-
liability of the Pressure Sore Status Tool. Decubitus 1992; 5:
REFERENCES
22–8.
1. Levine NS, Lindberg RB, Mason AD, Pruitt BA. The quanti- 15. Hanley JA, McNeil BJ. The meaning and use of area under a
tative swab culture and smear: a quick, simple method for de- receiver operating characteristic (ROC) curve. Radiology
termining the number of viable aerobic bacteria on open 1982; 143: 29–36.
wounds. J Trauma 1976; 16: 89–94. 16. DeLong ER, Delong DM, Clarke-Pearson DL. Comparing
2. Sapico FL, Ginunas VJ, Thornhill-Joynes M, Canawati HN, the areas under two or more correlated receiver operating
Capen DA, Klein NE, Khawan S, Montgomerie JZ. Quanti- characteristic curves: a nonparametric approach. Biometrics
tative microbiology of pressure sores in different stages of 1988; 44: 837–45.
healing. Diagn Microbiol Infect Dis 1986; 5: 31–8. 17. Pepe MS. The statistical evaluation of medical tests for classi-
3. Basak S, Dutta SK, Gupta S, Ganguly AC, De R. Bacteriol- fication and prediction. New York: Oxford University Press,
ogy of wound infection: evaluation by surface swab and quan- 2003.
titative full-thickness wound biopsy culture. J Ind Med Assoc 18. Bill TJ, Ratliff CR, Donovan AM, Knox LK, Morgan RF,
1992; 90: 33–4. Rodeheaver GT. Quantitative swab culture versus tissue
4. Herruzo-Cabrera R, Vizcaino-Alcaide MJ, Pinedo-Castillo C, biopsy: a comparison in chronic wounds. Ostomy/Wound
Rey-Calero J. Diagnosis of local infection of a burn by semi- Manage 2001; 47: 34–7.
quantitative culture of the eschar surface. J Burn Care Rehab 19. Gardner S, Frantz R, Doebbeling B. The validity of the clin-
1992; 13: 639–41. ical signs and symptoms used to identify localized chronic
5. Rudensky B, Lipschits M, Isaacsohn M, Sonnenblick M. In- wound infection. Wound Rep Reg 2001; 9: 178–86.
fected pressure sores: comparison of methods for bacterial 20. Bowler PG. The 10(5) bacterial growth guideline: reassessing
identification. Southern Med J 1992; 85: 901–3. its clinical relevance in wound healing. Ostomy/Wound
6. Krizek TJ, Robson MC. Evolution of quantitative bacteriolo- Manag 2003; 49: 44–53.
gy in wound management. Am J Surg 1975; 130: 579–84. 21. Johnson S, Lebahn F, Peterson LR, Gerding DN. Use of an
7. Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolkin RH. anaerobic collection and transport swab device to recover
Manual of clinical microbiology. 7th ed. Washington, DC: anaerobic bacteria from infected foot ulcers in diabetics. Clin
American Society for Microbiology, 1999. Infect Dis 1995; 20 (Suppl. 2): S289–90.
Wound Rep Reg (2006) 14 548–557 2006 by the Wound Healing Society
c 557