For patients with possible Staphylococcus aureus infection, providers must decide whether to treat empirically for methicillin-resistant S. aureus (MRSA). Nares MRSA colonization screening tests could inform decisions regarding empiric MRSA-active antibiotic use.Reference Smith, Erdman, Ferreira, Aldridge and Jankowski1, Reference Chotiprasitsakul, Tamma, Gadala and Cosgrove2
The negative predictive value (NPV) of MRSA nares tests has varied greatly among studies, Reference Smith, Erdman, Ferreira, Aldridge and Jankowski1–Reference Robicsek, Suseno, Beaumont, Thomson and Peterson9 from 45%Reference Schleyer, Jarman, Chan and Dellit3 to 99%.Reference Dangerfield, Chung, Webb and Seville4 Despite increasing attention to the tests’ potential clinical utility, especially for respiratory infections, Reference Parente, Cunha, Mylonakis and Timbrook10 factors that influence its NPV remain largely unstudied. Accordingly, we assessed multiple clinical variables as correlates of MRSA nares tests’ NPV for methicillin resistance among clinical S. aureus isolates.
Methods
This retrospective cohort study involved inpatients at the Minneapolis Veterans Affairs Medical Center (MVAMC), a 207-bed facility with 48 rehabilitation beds. According to a Veteran Affairs national directive, during the study period (2013–2016) patients underwent MRSA nares screening using either polymerase chain reaction (PCR, Xpert MRSA, Cepheid, Sunnyvale, CA) or culture (CHROMagar, Becton Dickinson, Franklin Lakes, NJ) on admission, discharge, and transfer (2013–2015), or only on admission and intensive care unit (ICU) transfer (2016). After institutional review board approval, the clinical cohort was assembled by querying a microbiology laboratory database to identify inpatient S. aureus isolates, excluding “stool,” “rectal,” “vaginal,” “genital,” “throat,” “nares,” and “nasal” as likely colonization, not infection.Reference Dangerfield, Chung, Webb and Seville4 Of 1,039 total S. aureus clinical isolates, 445 were represented by patient replicates; only the most recent isolate per patient was retained. Of 594 remaining unique isolates, 36 were excluded (35 lacked MRSA nares testing and 1 yielded both MRSA and MSSA). The final population comprised 558 patients (96% male; mean age, 66.7 years).
Oxacillin-resistant clinical isolates (ie, MRSA) identified using Vitek 2 (bioMèrieux, Marcy-l’Étoile, France) were categorized by time interval between nares screening and clinical isolate specimen collection (≤30 or >30 days), patient location (ICU, medical ward [medical], rehabilitation/spinal cord unit [rehabilitation]), specimen type (respiratory, blood, skin/soft tissue, urine, bone/joint, or urine), and nares test method (PCR or culture).
Primary outcomes included the sensitivity, specificity, positive predictive value (PPV), and NPV of nares tests for the MRSA status of the isolate’s status. Overall NPV was calculated by considering (1) only the most recent prior nares test and (2) all nares tests from 12 months prior to clinical isolate specimen collection. All other analyses used only the nares test immediately prior to isolate specimen collection. The Fisher exact test (2-tailed) was used for between-subcategory comparisons. Multivariable analysis was used to assess jointly patient location, time interval, and their interaction term.
Results
Of the 558 S. aureus clinical isolates, 38% were MRSA. The NPV of the nares tests for MRSA was 81.0% (95% confidence interval [CI], 78.1–83.6) when only the most recent nares test was considered, and this value increased negligibly to 82.6% (95% CI, 78.5–86.1) when any positive nares test in the prior 12 months was considered. By contrast, test performance varied substantially by clinical subgroup (Table 1).
Performance Characteristics of Methicillin-Resistant Staphylococcus aureus Nares Screens in Relation to Clinical Variables
Note. PPV, positive predictive value; NPV, negative predictive value; PCR, polymerase chain reaction.
a Time interval between collection of the index culture and collection of the nares screening test.
b Patient location at the time of the index culture collection.
c Specimen type of the index culture that yielded S. aureus.
d N/A, not applicable due to insufficient sample size.
e Method used for nares screening test (both PCR and culture were used during the study).
Time interval
Compared with nares tests collected >30 days before clinical isolate specimen collection (n = 86, 25.4%), tests collected at a shorter interval (≤30 days: n = 472, 84.6%) had a significantly higher NPV (82.8% [95% CI, 78.4–86.7] vs 72.2% [95% CI, 60.4–82.1]; P = .046).
Location
Nares tests from rehabilitation patients had significantly lower NPV (68.2% [95% CI, 60.9–74.7]) than those from the ICU (88.6% [95% CI, 79.0–94.2]; P = .02) or medical patients (82.7% [95% CI, 79.3–85.6]; P = .01).
Specimen
Test performance varied somewhat (albeit nonsignificantly) by specimen type. The NPV ranged from 75% (95% CI, 64.3–83.4) for urine to 84% (95% CI, 77.2–89.1) for bone and joint specimens.
PCR versus culture
The NPV was similar for PCR-based versus culture-based nares tests, i.e., 81.7% (95% CI, 78.0–84.8) versus 79.6% (95% CI, 74.8–83.6), respectively (P = 0.6).
Other subcategories
In a multivariate model that included location and time interval, location remained a significant predictor of MRSA status overall (P < .01), whereas time interval did not (P = .13). Stratification by location identified time interval as a significant predictor of NPV for the rehabilitation unit (P = .03) but not medical wards (P = .66) or the ICU (P = .46).
Discussion
This retrospective cohort study involving MVAMC inpatients assessed MRSA nares screening tests for predicting MRSA status of clinical S. aureus isolates. We identified a significantly higher nares-test NPV in 2 subgroups: clinical isolate specimens obtained from ICU or medical ward (vs rehabilitation) patients, and nares tests done within 30 days (vs >30 days) of isolate specimen collection.
Our study has demonstrated a higher MRSA proportion among clinical isolates (38%) than all but 2Reference Schleyer, Jarman, Chan and Dellit3, Reference Rimawi, Ramsey, Shah and Cook7 of 9 referenced studies.Reference Smith, Erdman, Ferreira, Aldridge and Jankowski1–Reference Robicsek, Suseno, Beaumont, Thomson and Peterson9 Because NPV varies inversely with targeted condition prevalence (here, MRSA), our comparatively low overall NPV may reflect our comparatively high overall MRSA prevalence. Likewise, the comparatively low nares screen NPV from the rehabilitation unit may reflect that unit’s higher MRSA fraction among S. aureus isolates (50.0%) compared with the ICU (28.1%) or medical wards (36.7%).
Our study has several limitations. First, because it was hypothesis generating, observed differences between subgroups require confirmation. Second, certain subgroups were small, limiting power. Third, the distinctive veteran population may limit generalizability. Fourth, we included all clinical isolates, regardless of whether they represented active infections. Finally, all subjects had an S. aureus clinical isolate. Inclusion of other subjects predictably would increase the test’s NPV due to the lower pretest probability of MRSA.
In summary, we found a significantly higher NPV for MRSA nares screening tests performed within 30 days of S. aureus clinical isolate specimen collection and in units with lower MRSA prevalence (ICU and medical, vs rehabilitation). Use of MRSA nares tests in empiric antibiotic selection may benefit from consideration of factors influencing test performance, including time since nares test and background MRSA prevalence.
Acknowledgments
We thank the MVAMC microbiology laboratory for their assistance, particularly with the creation of our data set.
Author ORCIDs
Teresa Fox, 0000-0003-2908-8387
Financial support
This work was supported in part by Office of Research and Development, Department of Veterans Affairs. The opinions expressed are strictly those of the authors and not those of the Department of Veterans Affairs or the MVAMC.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
