Clinical Impact of a Real-Time PCR Assay for Rapid Identification of ...

Clinical Impact of a Real-Time PCR Assay for Rapid Identification of Staphylococcal Bacteremia Abigail M. Frye,a Catherine A. Baker,a D. Leif Rustvold,b Kim A. Heath,c Jessica Hunt,a James E. Leggett,d,e and Margret Oethingerc Department of Pharmacy, Providence St. Vincent Medical Center, Portland, Oregon, USAa; Center for Outcomes Research and Education, Portland, Oregon, USAb; Department of Laboratory Servicesc and Department of Medical Education,d Providence Portland Medical Center, Portland, Oregon, USA; and Department of Medicine, Oregon Health and Science University, Portland, Oregon, USAe

The purpose of this study was to evaluate the impact of real-time PCR reporting both on timely identification of clustered Grampositive cocci (GPC) in blood cultures and on appropriate antibiotic treatment. This retrospective, interventional cohort study evaluated inpatients with blood cultures positive for GPC in the pre-PCR (15 January 2009 to 14 January 2010) and post-PCR (15 January 2010 to 14 January 2011) periods. Post-PCR implementation, laboratory services completed batched PCR; results other than methicillin-resistant Staphylococcus aureus (MRSA) were reported in the electronic medical record without additional interventions. The assay’s sensitivity and specificity, time to identification of staphylococcal bacteremia, and clinically relevant outcomes, including time to optimal antibiotic therapy, were evaluated. Demographic information was also collected and analyzed. Sixty-eight and 58 patients with Staphylococcus aureus bacteremia from the pre- and post-PCR periods, respectively, met inclusion criteria. Similar numbers of consecutive patients with coagulase-negative staphylococci were analyzed for comparison. The time to identification was significantly reduced post-PCR implementation (mean, 13.2 h; 95% confidence interval [95% CI], 10.5 to 15.9 h; P < 0.0001). However, the time to optimal antibiotic therapy was not significantly reduced. We conclude that implementation of a PCR assay demonstrated the potential to improve appropriate antibiotic use based on clinically meaningful and statistically significant reductions in the time to microbiologic identification. However, in order to realize this potential benefit, processes must be optimized and additional interventions initiated to facilitate providers’ use of the PCR result.

S

taphylococcus aureus bacteremia causes significant morbidity and mortality. Delay in the initiation of appropriate antimicrobial therapy has been identified as an important determinant of clinical outcomes (1, 6). As a result, rapid identification of staphylococcal species in bacteremic patients provides theoretical clinical advantages in the optimization of therapy. Pathogenic S. aureus must be differentiated from clinically insignificant coagulase-negative staphylococci (CoNS). Methicillin-resistant S. aureus (MRSA), whose prevalence is increasing and which accounts for 30 to 40% of all S. aureus infections, must be identified (8, 9). Delay in the administration of appropriate antibiotics for bacterial bloodstream infections caused by MRSA is associated with increased mortality, prolonged hospitalization, and secondary infections (e.g., endocarditis and vertebral osteomyelitis) (11, 14). Specifically, a delay in appropriate antibiotic treatment greater than 44 to 48 h after the collection of a blood culture positive for MRSA is associated with increased infection-related mortality and a prolonged length of hospital stay (11, 13). As a result, vancomycin—a glycopeptide antibiotic currently considered first-line antibiotic therapy for MRSA—is often started as empirical antibiotic therapy when clustered Gram-positive cocci (GPC) are identified in a blood culture. Unfortunately, vancomycin is not optimal therapy for methicillin-susceptible S. aureus (MSSA) bacteremia, for which ␤-lactam antibiotics with activity against staphylococci are superior (3, 7). Thus, rapid identification of staphylococcal species and methicillin susceptibility has important implications in guiding early optimal antibiotic therapy. Whereas conventional methods used to differentiate S. aureus from CoNS and MRSA from MSSA require approximately 24 to 48 h from the time blood cultures turn positive, a rapid PCR assay can provide species identification and the methicillin susceptibility or resistance of an S. aureus isolate within 2 to 6 h after a

0095-1137/12/$12.00

Journal of Clinical Microbiology

p. 127–133

positive blood culture is detected. In theory, if staphylococci growing in blood cultures can be identified faster, both the time to optimal antibiotic therapy and the time to discontinuation of unnecessary antibiotics (e.g., vancomycin) should be reduced. Moreover, indiscriminate use of vancomycin is associated with a risk of increased glycopeptide resistance in enterococci (14). During our preliminary validation study of the real-time Becton Dickinson (BD) GeneOhm StaphSR PCR assay (Becton Dickinson Diagnostics, Québec, QC, Canada), performed twice daily, the time to species identification was reduced by an average of 20 h. Fifty-four percent (13/24) of patients could have benefited from earlier optimization of antibiotic therapy, which would have reduced unnecessary vancomycin usage for 42% (10/24) of patients. These improvements were similar to the results of other investigators both when potential benefits were theoretically calculated and when PCR results were reported to clinicians (1, 12, 14). However, other interventions were usually introduced at the same time as the PCR assay, most notably changes in clinician notification or active antibiotic monitoring involving the pharmacy department. Hence, the impact of the PCR assay alone could not be accurately assessed. The current study was undertaken to evaluate the clinical impact of rapid PCR result reporting alone, without any other concomitant changes.

Received 20 October 2011 Accepted 21 October 2011 Published ahead of print 9 November 2011 Address correspondence to Abigail Frye, [email protected]. Copyright © 2012, American Society for Microbiology. All Rights Reserved. doi:10.1128/JCM.06169-11

jcm.asm.org

127

Frye et al.

TABLE 1 Definitions of optimal, suboptimal, and nonoptimal antibiotic therapya Organism

Optimal therapy

Suboptimal therapy

Nonoptimal therapy

MSSA

Antistaphylococcal ␤-lactam, such as oxacillin, extended-spectrum or broad-spectrum cephalosporin, ␤-lactam–␤-lactamase inhibitor; vancomycin (if the patient is ␤-lactam allergic or if this is otherwise determined to be optimal therapy by an ID physician) Vancomycin, daptomycin, linezolid

Other antibiotics to which the isolate is susceptible, but which are clinically inferior (e.g., vancomycin for patients without ␤-lactam allergy, oral ␤-lactams, etc.)

Antibiotics to which the isolate is not susceptible

Other antibiotics to which the isolate may be susceptible, but which are suboptimal (e.g., clindamycin, doxycycline, trimethoprim-sulfamethoxazole)

Antibiotics to which the isolate is not susceptible

MRSA

a

For CoNS, the appropriateness of continued antibiotic therapy (⬎48 h after positive culture) was determined by an ID physician.

MATERIALS AND METHODS This retrospective, interventional cohort study was conducted at two 500bed tertiary medical centers. The study was approved by the institutional review board and the privacy board, which waived informed consent. Laboratory services identified inpatients at the two medical centers who were 18 years of age or older with a blood culture positive for clustered GPC between 15 January 2009 and 14 January 2010 (referred to below as the “pre-PCR period”) and between 15 January 2010 and 14 January 2011 (the “post-PCR period”). Exclusion criteria were patient transfer with a previous positive blood culture and death or discharge within 48 h of the positive blood culture. If a patient had more than one positive blood culture during the same admission, PCR was run on the first culture only. The following laboratory data were collected: blood culture collection date and time, date and time of report of Gram stain, bacterial species identification, methicillin susceptibility, and final susceptibilities in the electronic medical record. The diagnostic workup was extracted from the laboratory information system (Cerner Millenium, Kansas City, MO). Demographic data and clinical outcome data were extracted from the electronic medical record; these included the age and sex of the patient, antibiotic allergies, date and time of administration of relevant antibiotics, date and time the physician discontinued suboptimal or unnecessary antibiotics, admission date and time, discharge date and time, and mortality data. Detailed analysis was completed on all inpatients within a 12-month period who had S. aureus isolated from their bloodstreams. In order to obtain a similarly sized subset of patients whose blood cultures yielded coagulase-negative staphylococci, we enrolled consecutive patients at one of the hospitals during a 10-month period both pre- and post-PCR implementation (January to October). All S. aureus isolates were considered clinically significant. For CoNS isolates, if antibiotics were never initiated or if antibiotics initiated in response to the positive Gram stain were stopped within 48 h after positive culture, the isolate was considered to be a contaminant and clinically insignificant. If antibiotics were continued for more than 48 h after a positive Gram stain, clinical significance was determined by an infectiousdisease (ID) physician (J.E.L.) on the basis of both laboratory data and patient-specific characteristics obtained by chart review. For all outcomes, T0 was defined as the time the blood culture was collected. Antibiotic therapy was defined as optimal, suboptimal, or nonoptimal on the basis of the species isolated and the organism’s susceptibility to methicillin (Table 1). Staphylococcus aureus isolates with discrepant PCR results were sent to Becton Dickinson Diagnostics–Infectious Diseases (Québec, Canada) for further molecular analysis. Laboratory methods. Throughout the study, the results of Gram staining of all positive blood cultures were called to the patient’s floor or the physician’s office within 60 min of the time when the blood culture bottle turned positive. Pre-PCR implementation, identification of S. au-

128

jcm.asm.org

reus was performed after 18 to 24 h of incubation by standard microbiological methods (typical colony morphology, Gram-positive cocci, positive catalase test, and positive latex coagulase test [Staphaurex test; Remel Europe Ltd., Kent, United Kingdom]). In addition, MRSA and MSSA were distinguished with a penicillin binding protein 2a (PBP2a) latex agglutination test (Oxoid Ltd., Basingstoke, Hants, United Kingdom) that takes about 20 min to perform. The PBP2a test was not performed on coagulase-negative staphylococci, because they have to be induced by antibiotics for 24 h, at which time antibiotic susceptibility information is usually available. MRSA identifications were called to the patient’s floor in order to ensure that appropriate contact precautions were initiated promptly. All other results (PBP2a-negative MSSA and coagulasenegative staphylococci) were simply released to the electronic medical record. The intervention consisted of implementation of the commercially available BD GeneOhm StaphSR PCR assay, which was performed in the regional microbiology laboratory. The PCR runs were batched twice daily, at 1000 and 1900 h, Monday through Friday and once daily, at 1500 h, on the weekend. If seeded cultures had grown for at least 18 h prior to the scheduled PCR run, the isolate was identified by the PBP2a test rather than by PCR. The BD GeneOhm StaphSR PCR assay targets a Staphylococcus aureus-specific gene for species identification and a sequence near the insertion site of the staphylococcal cassette chromosome mec element (SCCmec) for the identification of MRSA. If the S. aureusspecific target is present, the PCR result is designated “reactive,” indicating that S. aureus is present in the blood culture; if the mec flanking element is also detected, the PCR call is “positive,” indicating MRSA; if neither PCR target is detected with a positive internal control, the specimen is called “negative,” i.e., no S. aureus is detected. Clinicians were educated about the new process via the laboratory’s regular notification process, and beginning 15 January 2010, PCR results were released to the electronic medical record as follows: a “positive” result was reported as “presumptive methicillin-resistant Staphylococcus aureus,” a “reactive” result as “presumptive methicillin-susceptible Staphylococcus aureus,” and a “negative” result as “other than Staphylococcus aureus.” Importantly, clinician notification remained the same as before PCR implementation; MRSA results by PCR were called to the floor, but all other results were simply reported in the electronic medical record. No other interventions were introduced concurrently in order to assess the impact of the PCR intervention alone. Statistical methods. Student t tests or an analysis-of-variance (ANOVA) model were used to compare differences in means. Chi-square tests were used to compare in-hospital mortality rates. All statistical tests were computed for a two-sided type I error rate of 5%. The program “R: A Language and Environment for Statistical Computing, version 2.12” (R Foundation for Statistical Computing, Vienna, Austria) was used to perform statistical analysis. Outlier analysis was carried out on all samples using box plots; cases that fell outside the first or third quartile of the sample distribution by more than 1.5 times the interquartile range were


Clinical Impact of PCR for Staphylococcal Bacteremia

TABLE 2 Summary of patients with staphylococcal bacteremia in a 12-month period at two tertiary centers

TABLE 3 Patient demographics and clinical characteristics of all patients included in the study, before and after PCR implementation Valuea

No. of patients with: Staphylococcus aureus

Coagulase-negative staphylococci

Patients evaluated

Pre-PCR

Post-PCR

Pre-PCR

Post-PCR

Included in study Excluded from study Age, ⬍18 yr Death or discharge ⬍48 h after positive Gram stain Transferred from outside facility with previously positive blood culture Erroneously identified No data available in electronic medical record Duplicate isolates

68 164 3 10

58 118 4 6

66 80 4 23

52 66 11 25

14

18

2

1

2 1

1 2

0 3

0 3

134

87

48

26

Total

232

176

146

118

omitted. t tests of mean lengths of stay were performed on logarithmic transformations of days; all other tests were carried out on untransformed data. Pre- and post-PCR time to result means were compared first with ANOVA models controlling for potential confounders, including the medical center at which the specimen was collected, the bacterial strain, and patient age and gender. If significant confounding effects were identified, an ANOVA model was used in the final analysis of the comparison of means; otherwise, t tests were used.

RESULTS

Patients. During the 12-month pre-PCR period, laboratory services identified 232 S. aureus isolates in blood cultures collected from inpatients at the two large tertiary medical centers, of which 68 met inclusion criteria. During the 12-month post-PCR period, 176 S. aureus isolates were identified, of which 58 were included in the study. The reasons for exclusion (duplicate isolates in the majority of cases) are listed in Table 2. Sixty-six and 52 consecutive patients with CoNS isolated from blood cultures collected at one of the facilities during 10-month periods pre- and post-PCR implementation, respectively, met the inclusion criteria described. The patient demographics in the different groups were similar (Table 3). PCR performance. The regional microbiology laboratory serves 7 hospitals and numerous outpatient clinics. During the 12 months following PCR implementation, the sensitivity and specificity of the StaphSR PCR assay were evaluated. Compared to culture as the gold standard, the PCR assay was highly sensitive and specific (see Table A1 in the Appendix). Sensitivity for the identification of S. aureus versus non-S. aureus isolates was 99.4%, with a specificity of 99.8%. All MRSA isolates were correctly identified (sensitivity, 100%) (see the Appendix for further details). Post-PCR implementation, methicillin susceptibility was initially identified by PCR for 84% (49/58) of the S. aureus isolates in our study, by PBP2a for 14% (8/58), and by automated susceptibility testing for the remaining isolate (2%). Sixty-five percent (34/52) of the CoNS isolates were identified as non-S. aureus iso-

January 2012 Volume 50 Number 1

Parameter

Pre-PCR period

Post-PCR period

Mean age (yr) ⫾ SD Male sex Penicillin allergy Patients with CoNS Clinically significant Clinically insignificant Patients with S. aureus MRSA MSSA Length of stay (days) Mortality

64.0 (⫾15.0) 73 (54) 1 66 23 (35) 43 (65) 68 24 (35) 44 (65) 12.1 17 (12.7)

58.1 (⫾ 16.7) 68 (62) 3 52 14 (27) 38 (73) 58 25 (43) 33 (57) 11.6 14 (12.7)

a All values are numbers (percentages) of patients except where otherwise indicated. None of the differences were statistically significant.

lates by PCR, and 35% (18/52) were identified by the traditional latex coagulase test on isolated colonies. Impact of PCR implementation. (i) Time to identification. In the post-PCR period, the overall time to identification was reduced by 13.2 h (95% confidence interval [CI], 10.5 to 15.9 h; P ⬍ 0.0001). The average reduction in time to identification differed among MRSA, MSSA, and CoNS; however, the time to identification was reduced in both a clinically meaningful and a statistically significant (P ⱕ 0.004) manner for all three subgroups (Table 4). (ii) Time to optimal antibiotic therapy. (a) MRSA. The time to optimal antibiotic therapy for MRSA bacteremia did not improve after PCR implementation (Table 5). In fact, the average time to optimal antibiotic therapy was increased by an average of 3.7 h (P ⬎ 0.1). Notably, during the pre-PCR period, all patients with MRSA isolates were treated with vancomycin prior to the identification of MRSA. Conversely, in the post-PCR period, three MRSA isolates were not treated with optimal antibiotic therapy until MRSA was identified by PCR. (b) MSSA. Although the average time to identification of MSSA was reduced by 16.6 h (95% CI, 12.2 to 21.0; P ⬍ 0.0001), there was no significant difference in time to optimal antibiotic therapy. In the post-PCR period, the time to optimal antibiotic therapy was increased by 2.3 h, but this increase was not statistically significant. Fourteen MSSA isolates from the post-PCR period were initially treated with vancomycin as monotherapy (Table 5). In 2 of these cases, PCR was not performed. In 5 cases, the antibiotic therapy was optimized to a ␤-lactam antibiotic following the PCR result and prior to the final susceptibility data, and in 7 cases, antibiotic therapy was not optimized until after final susceptibilities were reported. For the 5 cases in which antibiotic therapy was optimized before final susceptibility results were available, there was a non-statistically significant trend toward a reduction in time to optimal antibiotic therapy (mean reduction, 6.9 h [P ⬎ 0.1]). Notably, in the 7 cases in which antibiotics were not optimized until after both PCR and susceptibility results were available, the time to optimal antibiotic therapy was significantly increased by an average of 21.4 h (95% CI, 3.0 to 33.7; P ⫽ 0.02). Five patients with staphylococcal bacteremia in the pre-PCR period and one patient with staphylococcal bacteremia in the post-PCR period were not included in the evaluation of the time to optimal antibiotic therapy, because they were already on opti-

jcm.asm.org 129

Frye et al.

TABLE 4 Time to presumptive identification of Gram-positive cocci in blood cultures of inpatients at two large tertiary care centers Pre-PCR

Post-PCR

Organism

No. of isolates (no. of outliers excluded)

Mean (range) time to identification (h)a


Mean (range) time to identification (h)a

Difference (h) (95% CI)

P

MRSA MSSA CoNS All

24 44 65 (1) 129 (5)

42.0 (26.4–70.1) 44.6 (24.5–70.5) 53.3 (30.4–85.4) 47.3 (24.5–77.4)

23 (2) 33 51 (1) 106 (4)

30.6 (22.4–55.3) 28.0 (16.9–47.2) 40.5 (20.1–82.5) 34.1 (16.9–66.0)

⫺11.4 (5.3–17.5) ⫺16.6 (12.2–21.0) ⫺12.8 (8.0–17.4) ⫺13.2 (10.5–15.9)

0.0004b ⬍0.0001b ⬍0.0001b ⬍0.0001c

a

Interval from the time when the blood culture was collected to the time of electronic reporting in the patient’s medical record. By t test. c By ANOVA with estimates from multivariate regression, controlling for species and methicillin susceptibility. b

mal antibiotic therapy prior to the time when the blood culture tested positive. CoNS. All patients with clinically significant CoNS isolates received optimal empirical antibiotic therapy. As a result, the impact of the PCR assay was not measured in time to optimal antibiotic therapy for clinically significant CoNS, but instead in time to discontinuation of unnecessary antibiotic therapy for clinically insignificant CoNS. Suboptimal and unnecessary vancomycin use. (i) MSSA. Of 44 patients with MSSA isolates during the pre-PCR period (Table 3), 38 patients (86%) received suboptimal vancomycin therapy. Two patients had documented ␤-lactam allergies, and therefore, the vancomycin was deemed appropriate; the remaining four patients did not receive vancomycin. In the post-PCR period, 76% (25/33) of patients with MSSA received suboptimal vancomycin therapy. Vancomycin therapy was considered appropriate for five patients due to documented ␤-lactam allergy or other clinical issues, and the remaining three patients did not receive vancomycin. Rapid PCR implementation was associated with an 8.7-h reduction in suboptimal vancomycin utilization (P ⫽ 0.09). During the post-PCR period, the time of suboptimal therapy differed

widely between patients stratified into different groups (Table 6). For 9 patients, suboptimal vancomycin was discontinued following the PCR result but before susceptibility results were available (a 19.6-h reduction). Nine other patients did not have the suboptimal vancomycin discontinued until after both the PCR and susceptibility results were available (a 4.0-h increase). One patient was switched to an optimal ␤-lactam before the PCR result was available (because MSSA was isolated from sputum), and the 6 remaining isolates were not identified by PCR. In the post-PCR period, 12 patients with MSSA bacteremia received vancomycin after the PCR result was available (mean, 2.5 doses; range, 1 to 7), but all patients were eventually treated with an antistaphylococcal ␤-lactam. (ii) CoNS. In the pre-PCR period, 25 of 66 patients (38%) with CoNS received unnecessary vancomycin. Vancomycin use was deemed appropriate by an infectious-disease physician (J.E.L.) for 23 patients, and the remaining 18 patients did not receive vancomycin. Similarly, 21 of 52 patients (40%) with CoNS isolates received unnecessary vancomycin in the post-PCR period. Sixteen patients received appropriate vancomycin therapy, and 15 patients did not receive vancomycin. For the evaluation of the time

TABLE 5 Time to optimal antibiotic therapy for inpatients with staphylococcal bacteremia at two large tertiary care centers Pre-PCR

Organism S. aureus MRSA MSSA MSSA isolates initially treated with vancomycin monotherapy All No PCR performed Antibiotics optimized after PCR, before C&Sf Antibiotics optimized after PCR and C&S

Post-PCR


Mean (range) time to optimal antibiotic therapy (h)a


Mean (range) time to optimal antibiotic therapy (h)a

Difference (h) (95% CI)

P

61 (2) 22 40 (1)

23.8 (0.2–74.7) 10.7 (0.22–24.72) 32.8 (0.27–96.1)

55 (2) 24 (1) 32

25.0 (0.5–91.3) 14.4 (0.67–36.88) 35.1 (0.52–98.48)

⫹1.2 (⫺5.1, 9.8) ⫹3.7 (⫺1.8, 9.1) ⫹2.3 (⫺10.5, 15.2)

⬎0.1b ⬎0.1c ⬎0.1d

15 (1)

55.3 (24.8–74.7)

14 2 5

62.3 (39.2–98.5) 57.3 (55.4,59.3) 48.4 (39.2–55.8)

⫹7.0 (⫺20.9, 24.9) ⫹2.0 ⫺6.9 (⫺21.6, 7.7)

⬎0.1e ⬎0.1 ⬎0.1d

7

73.7 (44.7–98.5)

⫹21.4 (3.0, 33.7)

0.02d

a

Interval between the time the blood culture was collected to the start of, or switch to, optimal antibiotic therapy (as defined in Table 1). By ANOVA controlling for the species, methicillin susceptibility, and the site. c By ANOVA controlling for the site. d By t test. e By ANOVA controlling for the timing of optimization. f C&S, culture and sensitivity. b

130

jcm.asm.org



TABLE 6 Impact of PCR implementation on suboptimal vancomycin use Pre-PCR Organism and parameter MSSA, duration of suboptimal vancomycin therapy All patients Vancomycin discontinued after PCR, before C&Se Vancomycin discontinued after PCR and C&S CoNS Duration of unnecessary vancomycin therapy Time to discontinuation of antibiotics for clinically insignificant CoNS

Post-PCR


Mean h (range)

37 (1)

46.8 (17.1–87.1)d

No. of isolates

Mean h (range)

Difference (95% CI)

Pb

25 9c

38.1 (10.7–82.1)d 27.2 (11.6–48.8)d

⫺8.7 (⫺19.0, 1.5) ⫺19.62 (⫺32.8, ⫺6.4)

0.09 0.0007

9c

50.8 (21.2–75.0)d

⫹4.0 (⫺10.0, 19.9)

⬎0.1

23 (2)

35.4 (14.8–67.7)d

21

34.3 (11.1–58.6)d

⫺1.1 (⫺10.1, 7.8)

⬎0.1

23 (3)

60.2 (32.8–89.6)a

23

57.4 (35.5–87.0)a

⫺2.8 (⫺11.1, 5.5)

⬎0.1

a

Interval from the time the blood culture was collected to the start of optimal antibiotic therapy or the discontinuation of unnecessary vancomycin. By t test. c For six patients, no PCR result was available; for one patient, vancomycin was discontinued before the PCR result was available. d Interval from the first dose of vancomycin to time of discontinuation. e C&S, culture and sensitivity. b

to discontinuation of antibiotics for clinically insignificant CoNS (Table 6), one patient during the pre-PCR period and two patients during the post-PCR period who received linezolid or daptomycin for clinically insignificant CoNS were also included in the analysis. The times to discontinuation of vancomycin for clinically insignificant CoNS were similar before and after PCR implementation (60.2 h versus 57.4 h; P ⬎ 0.1) (Table 6). Although the time to identification of CoNS was reduced by 12.8 h (P ⬍ 0.0001) following PCR implementation, this did not translate into a reduction in the duration of vancomycin therapy for these patients (35.4 h pre-PCR versus 34.3 h post-PCR; P ⬎ 0.1). Notably, 12 of the clinically insignificant CoNS isolates that were treated with vancomycin were correctly identified by PCR as “other than Staphylococcus aureus”; however, vancomycin was discontinued before sensitivity results were available in only 3 of these 12 cases. Ten of 21 patients with clinically insignificant CoNS isolates received unnecessary doses of vancomycin after the PCR result was available (mean, 2.2 doses; range, 1 to 3). There was no significant difference in length of stay or mortality between the periods before and after PCR implementation (Table 3; P, ⬎0.1 for all comparisons). DISCUSSION

By significantly reducing the time to identification, the implementation of a rapid PCR assay for the identification of staphylococcal bacteremia has confirmed the potential both to improve the time to optimal antibiotic therapy for MSSA and to reduce unnecessary vancomycin therapy in MSSA and clinically insignificant CoNS infections. However, the potential for reducing the overall time to optimal antibiotic therapy and for reducing unnecessary vancomycin utilization is realized only when the PCR result is acted upon promptly by the physician (2). Three other studies have recently reported the utility of a rapid PCR assay for the identification of staphylococcal bacteremia (1, 12, 14). Ruimy and colleagues compared the ideal antibiotic treatment potentially initiated if a PCR assay result were immediately available with the actual treatments chosen on the basis of pheno-


typic results (14). The investigators reported that 32% (39/121) of patients with clinically significant bacteremia (26 with MSSA, 5 with MRSA, 7 with methicillin-susceptible CoNS [MS-CoNS], and 1 with methicillin-resistant CoNS [MR-CoNS]) had received suboptimal antibiotic therapy and suggested that a PCR assay could have resulted in earlier optimization of therapy. Additionally, the investigators concluded that the use of a PCR assay could have prevented the unnecessary treatment of 3 patients with CoNS in only 1 of 2 blood cultures. These investigators assumed that the PCR assay result was available within 90 min of the Gram stain result, whereas in the current study, the PCR assays were batched and completed twice daily on weekdays and once daily on the weekend. As a result, the average interval between a positive Gram stain and the PCR result was 13 h. It is noteworthy that despite the batching of the PCR assay, we still achieved highly significant reductions in the time to identification. Nguyen and colleagues, evaluating vancomycin utilization for patients with methicillin-susceptible staphylococcal infections (both MSSA and MS-CoNS from a variety of sources, including blood, wound, sputum, and urine), reported that PCR implementation was associated with a median reduction in vancomycin use from 3 days (range, 1 to 44 days) to 1 day (range, 0 to 18 days) (P ⬍ 0.0001) and a median reduction in the time to antibiotic optimization from 5 days to 2 days (P ⬍ 0.0001) (12). They also reported that the use of a PCR assay was associated with more appropriate antibiotic use (38% pre-PCR versus 62% post-PCR), even though many patients (52% pre-PCR and 27% post-PCR) were never switched to alternative therapy after a finding of methicillin susceptibility. In our study, all patients with MSSA bacteremia, both pre- and post-PCR implementation, were eventually switched to optimal ␤-lactam therapy. Although Nguyen and colleagues reported that PCR implementation was also associated with a significant median reduction in the length of stay from 8 days prePCR (range, 1 to 47 days) to 5 days post-PCR (range, 0 to 42 days) (P ⫽ 0.03), the time to identification of MSSA was improved by 1 day, whereas the duration of vancomycin use was reduced by 2 days. This discrepancy suggests that factors in addition to PCR

jcm.asm.org 131

Frye et al.

implementation likely contributed to both the reduced vancomycin usage and the reduced length of hospitalization observed in their study. Of note, like our study, Nguyen’s study relied on routine result reporting in the electronic medical record without associated changes in reporting algorithms. However, a key difference in our study was the use of a rapid PBP2a test on blood culture isolates in the pre-PCR period that allowed presumptive MRSA identification about 18 to 24 h before conventional susceptibility testing. This may have decreased the impact of PCR implementation on the time to identification of MRSA in our study. Finally, Bauer and colleagues paired the implementation of a rapid PCR assay performed 24 h a day, 7 days a week, with an antimicrobial stewardship program that included an infectiousdisease (ID) pharmacist intervention (1). The PCR result was paged both to a specially trained ID pharmacist Monday to Friday, 0800 to 1700, and to the responsible physician 24 h a day, 7 days a week. Based on the PCR result, the ID pharmacist contacted the physician regarding any recommendations for alternative therapy. Post-PCR implementation, the time to antibiotic switch for patients with MSSA bacteremia was reduced by 1.6 days (P ⫽ 0.002). In this study, investigators did not include CoNS isolates. In contrast to the other available evidence regarding the use of a PCR assay for the identification of staphylococcal bacteremia, our study included consecutive staphylococcal isolates collected in the post-PCR implementation period as opposed to only those isolates identified by PCR. In addition, no other interventions were made concurrently with PCR implementation. Therefore, our study provides information regarding the realistic clinical impact that can be expected from implementing a PCR assay for rapid identification of GPC in blood cultures alone. The results from our study confirmed that the use of a rapid PCR assay alone is not likely to affect the time to appropriate antibiotic therapy for MRSA bacteremia. This is not surprising, given that vancomycin is now routinely used as empirical antibiotic therapy for patients with GPC in clusters on a Gram stain in light of the increasing prevalence of MRSA. As a result, the greatest impact of a rapid PCR assay for the identification of staphylococci in blood cultures is likely to be seen in the time to optimal antibiotic therapy for MSSA bacteremia and the time to discontinuation of unnecessary vancomycin therapy for MSSA bacteremia or clinically insignificant CoNS. However, the potential impact of more rapid identification will be realized only if the PCR result is acted upon promptly. To that end, performance of the PCR test 24 h a day, 7 days a week, is desirable; however, currently available random-access testing platforms, such as the GeneXpert system (Cepheid, Sunnyvale, CA), are cost-prohibitive for many institutions. We could show in our study that the time to identification was significantly reduced even though the PCR assay was batched twice a day. This did not translate into more-rapid optimization of therapy. Our study showed that the time to optimal antibiotic therapy and the time to discontinuation of unnecessary vancomycin use were reduced only in the small number of cases in which antibiotic therapy was optimized following the PCR result but before susceptibility data were available. Based on our results, it is clear that the use of a rapid PCR assay for the identification of staphylococci in blood cultures will improve clinically relevant outcomes only if many associated processes are optimized at the same time to facilitate timely physician response. Associated processes that need to be considered include

132

jcm.asm.org

the following: timing of PCR run times, method of clinician notification, education of non-infectious-disease physicians as to the benefits of ␤-lactam antibiotic therapy for methicillin-susceptible staphylococci, and knowledge of the sensitivity and specificity of the PCR assay among clinicians. In the current study, PCR results were typically reported in the electronic medical record either in the afternoon or between 2200 and midnight. Physicians may be more likely to respond to a PCR assay result made available earlier in the morning, when they are typically preparing for daily rounds. The recently published study by Bauer and colleagues provides evidence that physicians may be more likely to respond to a PCR result if they are actively notified of the result (as opposed to passive reporting of results in the electronic medical record) (1). Finally, providers may be more likely to respond to the PCR result if they are better educated regarding the specificity and sensitivity of the test. When we introduced the PCR assay in January 2010, our providers were informed of the new process by our customary methods, but we could not yet issue any local performance data in conjunction with the PCR result that was reported in the electronic medical record. In our study, there was a trend toward more providers optimizing therapy before final susceptibility data were available in the latter part of the year, suggesting that providers became more comfortable relying on the PCR assay as they learned more about it. Anecdotally, infectious-disease physicians were more likely to respond to the PCR assay, suggesting that physicians who were more familiar with molecular diagnostic tests were more likely to use the result to guide antibiotic therapy decisions. In our view, the effect of infectious-disease consultation should not be underestimated: in previous studies, involvement of infectious-disease physicians was associated with reduced mortality in patients with S. aureus bacteremia in the absence of rapid diagnosis (6, 10). Furthermore, Forrest et al. have reported outcome improvements for patients with S. aureus bacteremia from the use of a rapid fluorescence in situ hybridization (FISH) test (peptide nucleic acid [PNA]-FISH) that distinguished S. aureus from coagulase-negative staphylococci (4). They used an integrative approach, combining this technology with an antibiotic stewardship program, including active intervention by an infectiousdisease pharmacist if therapy needed to be changed, constant communication with providers, and frequent education of new residents and fellows. Interestingly, another study on the clinical impact of PNA-FISH testing of blood cultures with GPC, by Holtzman et al., failed to show a reduction in the length of hospital stay or in vancomycin usage in the absence of an antibiotic stewardship program (5). These data agree with our findings. In conclusion, the rapid PCR assay utilized in this study demonstrated high sensitivity and specificity compared with traditional phenotypic methods. The results of our study demonstrate that the use of a rapid PCR assay is associated with a statistically significant reduction in the time to identification of staphylococci in blood cultures, even when the PCR is run in a batched format. Our results also provide an important message regarding the implementation of rapid testing methods: “If you do it, do it right.” The implementation of a rapid PCR assay alone is not enough to improve antibiotic use for staphylococcal bacteremia. Associated processes, such as the timing of batched laboratory analysis, meaningful reporting of results, and—most desirable—the integration of rapid diagnosis into an antibiotic stewardship program, need to be optimized in order to fully impact timely patient care decisions at the bedside.



TABLE A1 Sensitivity and specificity of the GeneOhm StaphSR PCR assay performed during the 12-month study period, starting with the date of PCR implementation in routine clinical usea Organism (n)

Sensitivity (%)

Specificity (%)

Staphylococcus aureus (160) MRSA (56) MSSA (104) Gram-positive cocci other than S. aureus (520)

100 99.4 99.8

98.7 99.8 99.4

a The regional Clinical Microbiology Laboratory serves seven acute care hospitals and multiple physician offices.

APPENDIX

The regional microbiology laboratory serves 7 hospitals and numerous outpatient clinics. During the 12 months following the implementation of the BD GeneOhm StaphSR PCR assay, 680 PCR tests were performed on blood cultures that were positive for Grampositive cocci in clusters. The PCR assay was highly sensitive and specific compared to culture as the gold standard. Sensitivity for the identification of S. aureus versus non-S. aureus isolates was 99.4%, with a specificity of 99.8% (Table A1). All 56 MRSA isolates were correctly identified (sensitivity, 100%). Seven MSSA isolates, however, were initially identified as MRSA by PCR, resulting in a specificity of 98.7% for MRSA: one misidentification was due to technical error, and the other strains were subsequently identified as “reverter” strains, meaning that they had lost the functional mecA gene required for methicillin resistance but had retained residual staphylococcal mecA cassette (SCCmec) sequences targeted by the StaphSR PCR assay. The reverter strains identified were diverse and were not epidemiologically linked. One of the coagulase-negative staphylococci was misidentified as MRSA, but 519 isolates were correctly called “non-S. aureus” (sensitivity, 99.8%). Coagulase-negative staphylococci constituted the majority of the latter group (96.7%), followed by micrococci (1.3%) and a few isolates of six other Gram-positive genera. ACKNOWLEDGMENTS Becton Dickinson Diagnostics kindly provided molecular analysis of all staphylococcal isolates with discrepant PCR results. We thank Sue Corkum and Ivy Thomas for providing data for statistical analysis. The excellent technical expertise of the entire Microbiology PCR team is greatly appreciated. Dave Gilbert kindly reviewed the manuscript. There was no financial support for this study.


REFERENCES 1. Bauer KA, et al. 2010. An antimicrobial stewardship program’s impact with rapid polymerase chain reaction methicillin-resistant Staphylococcus aureus/S. aureus blood culture test in patients with S. aureus bacteremia. Clin. Infect. Dis. 51:1074 –1080. 2. Cattoir V, et al. 2011. Clinical impact of a real-time PCR assay for rapid identification of Staphylococcus aureus and determination of methicillin resistance from positive blood cultures. Clin. Microbiol. Infect. 17: 425– 431. 3. Chang FY, et al. 2003. Staphylococcus aureus bacteremia: recurrence and the impact of antibiotic treatment in a prospective multicenter study. Medicine 82:333–339. 4. Forrest GN, et al. 2006. Impact of rapid in situ hybridization testing on coagulase-negative staphylococci positive blood cultures. J. Antimicrob. Chemother. 58:154 –158. 5. Holtzman C, Whitney D, Barlam T, Miller NS. 2011. Assessment of impact of peptide nucleic acid fluorescence in situ hybridization for rapid identification of coagulase-negative staphylococci in the absence of antimicrobial stewardship intervention. J. Clin. Microbiol. 49: 1581–1582. 6. Honda H, Krauss MJ, Jones JC, Olsen MA, Warren DK. 2010. The value of infectious diseases consultation in Staphylococcus aureus bacteremia. Am. J. Med. 123:631– 637. 7. Kim SH, et al. 2008. Outcome of vancomycin treatment in patients with methicillin-susceptible Staphylococcus aureus bacteremia. Antimicrob. Agents Chemother. 52:192–197. 8. Klein E, Smith DL, Laxminarayan R. 2007. Hospitalizations and deaths caused by methicillin-resistant Staphylococcus aureus, United States, 1999 –2005. Emerg. Infect. Dis. 13:1840 –1846. 9. Klevens RM, et al. 2006. Changes in the epidemiology of methicillinresistant Staphylococcus aureus in intensive care units in US hospitals, 1992–2003. Clin. Infect. Dis. 42:389 –391. 10. Lahey T, Shah R, Gittzus J, Schwartzman J, Kirkland K. 2009. Infectious Diseases consultation lowers mortality from Staphylococcus aureus bacteremia. Medicine 88:263–267. 11. Lodise TP, McKinnon PS, Swiderski L, Rybak MJ. 2003. Outcomes analysis of delayed antibiotic treatment for hospital-acquired Staphylococcus aureus bacteremia. Clin. Infect. Dis. 36:1418 –1423. 12. Nguyen DT, et al. 2010. Real-time PCR testing for mecA reduces vancomycin usage and length of hospitalization for patients infected with methicillin-sensitive staphylococci. J. Clin. Microbiol. 48: 785–790. 13. Paul M, et al. 2010. Importance of appropriate empirical antibiotic therapy for methicillin-resistant Staphylococcus aureus bacteraemia. J. Antimicrob. Chemother. 65:2658 –2665. 14. Ruimy R, et al. 2008. Accuracy and potential usefulness of triplex realtime PCR for improving antibiotic treatment of patients with blood cultures showing clustered Gram-positive cocci on direct smears. J. Clin. Microbiol. 46:2045–2051.

jcm.asm.org 133