Utility of Broad-Range PCR Sequencing for Infectious Diseases Clinical Decision Making: a Pediatric Center Experience

ABSTRACT

Broad-range PCR (BRPCR) sequencing is a promising tool for diagnosis of infectious conditions when traditional microbiologic strategies fail to identify a pathogen. Data on the optimal clinical scenarios in which to use this tool are limited. We assessed, via retrospective chart review, the rate of organism identification and impact on clinical management from BRPCR testing sent from our quaternary care children’s hospital between February 2010 and June 2020. A total of 382 samples were sent from 269 individual patients. A total of 200 (74.3%) patients were immunocompromised. Median age at time of sample collection was 10.0 years (interquartile range, 4.2 to 15.8). A total of 254/377 (64.7%) samples were from patients known to be on ≥1 antimicrobial in the 24 h prior to sample collection. A total of 112/382 (29.3%) samples were from patients ultimately diagnosed with a bacterial or fungal infection by another testing modality. The most common sample types were bronchoalveolar lavage (BAL) fluid (45), lung tissue (41), and bone (39). An organism was identified from 83 (21.7%) samples, but results from only 19 (5.0%) samples led to a change in management. Organisms were identified from 18 (40%) BAL samples; only 2 (4.4%) were judged to be clinically significant. A total of 4/12 (33.3%) samples from cardiac hardware changed clinical management. We found that only 5% of BRPCR results influenced antimicrobial management in a diverse pediatric cohort. Our findings suggest that the impact on clinical management varied widely by sample type. Additional work is necessary to characterize the ideal clinical scenarios in which BRPCR should be used.

INTRODUCTION

Broad-range PCR (BRPCR) testing consists of gene amplification and sequencing with results compared to validated databases for organism identification, with differing targets depending on organism type. This technique has several theoretical benefits over conventional microbiologic studies, including the option to add on testing to saved specimens after an initial round of testing (1). It also allows providers to perform hypothesis-generating rather than hypothesis-confirming testing, especially when conventional testing has been unsuccessful in identifying an organism. However, BRPCR has several limitations. Its use is limited to detection of bacteria and fungi, so alternative testing must be included if viral or parasitic infections are on the differential. It is unable to provide susceptibility data, so it is unlikely to entirely replace culture-based diagnostic methods in routine evaluations (2). This test is also more expensive than many traditional methods (2, 3), and it has significantly lower sensitivity than PCRs specific to an organism (4, 5). Furthermore, similar to other tests that are not targeted to a specific pathogen, BRPCR has the potential to identify colonizing or contaminating organisms. The results must therefore be carefully considered to determine whether they represent true infection (3, 6).

Data to support optimal use scenarios for BRPCR remain limited, especially in pediatrics. It is generally considered particularly beneficial for identification of fastidious or slow-growing organisms. As up to 20% cases of endocarditis are culture negative (7), there has been significant attention to the use of BRPCR in this clinical scenario (8–12). Similarly, it has been considered for use in osteoarticular infections, as some known causative organisms, such as Kingella kingae, can be difficult to identify with traditional microbiologic methods (13, 14).

Given the limited data available for pediatric patients, we undertook a review of the BRPCR samples sent from our pediatric center over the past decade to characterize practice patterns and provide additional data about sample types that are most likely to change clinical management.

MATERIALS AND METHODS

We completed a retrospective chart review of patients with BRPCR testing sent from our quaternary freestanding children’s hospital between February 2010 and June 2020. At our institution, BRPCR from bronchoalveolar lavage (BAL) fluid requires approval from the Infectious Diseases Diagnostic Laboratory directors; no approval is required for any other sample type. Most, but not all, decisions to send samples for BRPCR are made in conjunction with the infectious diseases consult service. All samples are sent to the University of Washington Molecular Microbiology Laboratory for this testing, and providers complete a requisition form in which they select among bacterial, fungal, and AFB BRPCRs. At the University of Washington, sequencing of the 16S rRNA gene is used for bacteria; 28S rRNA and internal transcribed spacer regions are used for fungi; and multiple strategies, including targeting the 65-kilodalton heat shock protein, are used for acid-fast bacilli (AFB) (15). The requisition form provides the option to select pathogen-specific PCRs and to indicate whether to reflex to next-generation sequencing if multiple templates are identified (16).

To identify cases, we used a list provided by the Infectious Diseases Diagnostic Laboratory of all samples sent to the University of Washington for BRPCR testing during the study period. We excluded samples if they were sent for only targeted, not broad-range, PCRs. We also excluded samples if the scanned result report could not be located in the electronic medical record. We included samples regardless of patient age; our hospital, like many quaternary pediatric institutions, cares for a sizable population of adults with conditions originating in childhood. All eligible records were reviewed by the lead author (C.N.L.); data were recorded in a secure REDCap (Research Electronic Data Capture, Vanderbilt, TN) database (17).

From chart review, we collected data on patient demographics; clinical presentation, including immune status; sample details; type of BRPCR sequencing; additional infectious diseases work-up, including pathology evaluation; and antimicrobial treatment. We did not set a strict cutoff time before and after BRPCR sampling for inclusion of additional infectious diseases work-up; we instead determined which tests were sent to evaluate the same clinical presentation based on review of clinician notes. We determined that a BRPCR result had changed clinical management if ≥1 antimicrobial was started or stopped due to the result. If the clinical notes did not specifically state that the change was made due to the BRPCR result, the authors made a judgment based on the timing of that result relative to documented management changes. We reviewed other microbiologic testing sent as part of the diagnostic evaluation and clinician notes to determine whether an infectious disease diagnosis was made by another modality. Examples of this testing included but were not limited to pathogen-specific PCRs, serologic testing, and conventional cultures. Pathology results were considered positive if bacterial or fungal forms were mentioned in the final pathology report; we did not consider positive viral stains (e.g., Epstein-Barr virus [EBV]-encoded small RNA staining) on pathology specimens to be positive for the purposes of this analysis. Preservation status of samples (e.g., fresh versus paraffin-embedded) was not routinely documented in the medical record and thus could not be evaluated as a possible factor affecting BRPCR yield. It was not possible to collect the exact turnaround time, which could have affected the impact of results on management, for all specimens, so time from sample collection to receipt at the reference laboratory was used as a surrogate measure.

Data were analyzed using descriptive statistics. Analyses were performed in JMP Pro 15.0.0 (SAS Institute, Cary, NC) and Stata BE 17.0 (StataCorp, College Station, TX). We used 2-tailed Fisher’s exact tests for comparisons; when comparing sample types, only those with at least 5 samples were included in the statistical analysis. We considered a P value of <0.05 statistically significant.

The Boston Children’s Hospital Institutional Review Board reviewed the study protocol and determined that it qualified as exempt from the requirements of 45 CFR 46.

RESULTS

A total of 382 BRPCR samples were identified from 269 unique patients: 302 samples were sent for bacterial sequencing, 286 for fungal, and 305 for acid-fast bacilli (AFB). The median patient age at the time of sample collection was 10.0 years (interquartile range [IQR], 4.2 to 15.8). Two hundred (74.3%) patients had compromised immune systems due to either primary or acquired immunodeficiency or treatment with immunosuppressive therapies (Table 1). A definitive diagnosis of a bacterial or fungal infection was made by a modality other than BRPCR in 112/382 (29.3%) cases. Overall, 83/382 (21.7%) BRPCR samples identified at least one organism. A total of 19 of 382 (5.0%) samples resulted in a change in clinical management, 13 (3.4%) due to identification of an organism and 6 (1.6%) due to the absence of an identified organism. A total of 8 samples resulted in discontinuation of antimicrobials, 6 in addition of antimicrobials, and 5 in transition between antimicrobial regimens (Tables 2 and 3).

TABLE 1.

Patient demographics

Parameter	Value (%)
Age (yr; median, IQR)	10.0 (4.3–15.7)
Female	157 (58.4)
Immunosuppressed	200 (74.3)
Primary immunodeficiency	63 (31.5)
History of HSCTa	43 (21.5)
Active chemotherapy	40 (20)
On treatment with immunosuppressives	29 (14.5)
Solid-organ transplant	23 (11.5)
Aplastic anemia	15 (7.5)
HIV	1 (0.5)
Samples with a bacterial or fungal organism identified on pathology	67/313 (21.4)
Samples in which patient was on antimicrobials in 24 hours prior to sample collection	254/377 (67.4)

^aHSCT, hematopoietic stem cell transplant.

TABLE 2.

Sample characteristics^d

Parameter	Organism identified, impact on antimicrobial management (n = 13)	Organism identified, no impact on antimicrobial management (n = 70)	No organism identified, impact on antimicrobial management (n = 6)	No organism identified, no impact on antimicrobial management (n = 293)
Receipt of antimicrobials during 24 h prior to sample collection,a n (%)	9 (3.5)	45 (data not available for 1) (17.7)	5 (2.0)	195 (data not available for 4) (76.8)
Infectious disease diagnosed by other modality, n (%)	3 (2.7)	34 (30.4)	2 (1.8)	73 (65.2)
Sample type, n (%)
BALb	2 (4.4)	16 (35.6)	0 (0)	27 (60.0)
Lung	2 (4.9)	2 (4.9)	1 (2.4)	36 (87.8)
Bone	0 (0)	6 (15.4)	1 (2.6)	32 (82.1)
Lymph node	0 (0)	6 (19.4)	0 (0)	25 (80.6)
CSFc	0 (0)	3 (10.7)	0 (0)	25 (89.3)
Abscess	1 (4.0)	6 (24.0)	0 (0)	18 (72.0)
Intestinal tissue	0 (0)	6 (28.6)	0 (0)	15 (71.4)
Soft tissue	1 (5.0)	3 (15.0)	0 (0)	16 (80.0)
Liver tissue	1 (5.3)	4 (21.1)	1 (5.3)	13 (68.4)
Pleural fluid	1 (6.7)	5 (33.3)	0 (0)	9 (60.0)
Cardiac Hardware	2 (16.7)	1 (8.3)	2 (16.7)	7 (58.3)
Cardiac tissue	1 (10.0)	1 (10.0)	0 (0)	8 (80.0)
Skin biopsy	1 (11.1)	1 (11.1)	0 (0)	7 (77.8)
Wound	0 (0)	2 (28.6)	0 (0)	5 (71.4)
Joint fluid	0 (0)	0 (0)	0 (0)	6 (100.0)
Bone marrow	0 (0)	0 (0)	0 (0)	6 (100.0)
Synovial tissue	0 (0)	1 (20.0)	0 (0)	4 (80.0)
Pericardial fluid	0 (0)	0 (0)	0 (0)	4 (100.0)
Peritoneal fluid	0 (0)	1 (25.0)	0 (0)	3 (75.0)
Brain tissue	0 (0)	0 (0)	0 (0)	3 (100.0)
Other	1 (3.1)	6 (18.8)	1 (3.1)	24 (75.0)
Total	13 (3.4)	70 (18.3)	6 (1.6)	293 (76.7)

^aFive samples were obtained outside our institution, and it is not known whether the patient was on antimicrobials at the time of collection.

^bOf the 16 BAL samples in which an organism was identified but there was no change in clinical management, 6 were from patients already on appropriate therapy; 6 had multiple organisms identified, and the results were felt to represent colonizing airway flora; 4 were from patients determined to have a noninfectious cause of lung disease (2 also with multiple organisms identified); 1 had an identified organism that was felt not to explain the clinical picture; and 1 was tested to rule out a specific organism, but a different organism was incidentally identified and not felt to be clinically significant.

^cCSF, cerebrospinal fluid.

^dAll percentages were calculated as row percentages.

TABLE 3.

Clinical details of specimens with BRPCR results leading to a change in clinical management^a

Case	Clinical scenario	Immunocompromised	Sample type	Time from collection to sample receipt at UW (days)	Type of broad-range sequencing done	Organism(s) identified	Additional microbiologic testing	Additional microbiologic testing positive	Organisms seen on pathology	Final diagnosis	Management change
1	Shoulder pain following gunshot wound	No	Abscess	3	Bacterial, mycobacterial, fungal	Fusobacterium necrophorum	Conventional culture	No	Not evaluated	Fusobacterium necrophorum abscess and osteomyelitis	Vancomycin discontinued, metronidazole started
2	Septic joint and cardiac vegetation in patient s/p surgery for CCHD	No	Cardiac hardware	8	Bacterial, mycobacterial, fungal	Streptococcus gordonii	Conventional culture, AFB culture, targeted serologies, next- generation sequencing	Yes, blood culture prior to transfer grew Streptococcus gordonii; next-generation sequencing detected Escherichia coli and Haemophilus influenzae	No	Streptococcus gordonii mitral valve endocarditis	Vancomycin and ampicillin-sulbactam discontinued, ceftriaxone started
3	Fever and pulmonary nodules	Yes, history of chronic granulomatous disease s/p HSCT	Lung tissue	1	Bacterial, mycobacterial, fungal	Aspergillus fumigatus	Conventional culture, AFB culture, targeted PCR, targeted serologies, galactomannan, 1,3 β-d-glucan, targeted antigens for endemic fungi, viral culture, viral DFA	Yes, 1,3 β-d-glucan moderately elevated	No	Invasive Aspergillus fumigatus	Meropenem and liposomal amphotericin B discontinued, voriconazole started
4	Baclofen pump infection	No	Baclofen pump fluid collection	5	Bacterial, mycobacterial, fungal	Streptococcus pyogenes	Conventional culture	No	Not evaluated	Streptococcus pyogenes baclofen pump infection	Vancomycin and cefepime discontinued, ceftriaxone started
5	Inability to wean off steroids in setting of bronchiolitis obliterans	No	BAL	14	Bacterial, mycobacterial, fungal	Tropheryma whipplei (detected with both broad- range and targeted PCR)	Conventional culture, targeted PCR, IGRA, viral culture, PJP stain	Yes, AFB stain with <1 per field, fungal culture grew Streptomyces, lung biopsy with Staphylococcus species not aureus and Staphylococcus warneri	No	Tropheryma whipplei infection superimposed on underlying lung disease	Trimethoprim-sulfamethoxazole started
6	Fever, cough, pulmonary nodules	No	BAL	11	Bacterial, mycobacterial, fungal	Cutibacterium acnes	Conventional culture, AFB culture, targeted PCRs, IGRA, galactomannan, 1,3 β-d-glucan, cytomegalovirus shell vial culture, PJP stain, viral DFA, antigens for endemic fungi	No	No	Pulmonary nodules attributed to Cutibacterium acnes	Cefdinir started
7	Liver lesions	Yes, SCID	Liver tissue	Unknown	Bacterial, mycobacterial, fungal	Legionella pneumophila	Conventional culture, AFB culture, targeted PCRs, targeted serologies, galactomannan, 1,3 β-d-glucan	No	No	Liver abscesses secondary to Legionella pneumophilia	Ciprofloxacin started
8	Fevers	No	Skin biopsy	6	Bacterial, mycobacterial, fungal	Achromobacter xylosoxidans	Conventional culture, AFB culture, targeted PCRs, targeted serologies, PPD, IGRA, galactomannan, 1,3 β-d-glucan, antigens for endemic fungi	No	No	Presumed Achromobacter xylosoxidans infection	Piperacillin-tazobactam started
9	Leg lesion	Yes, X-linked agammaglobulinemia	Soft tissue	Unknown	Bacterial, mycobacterial, fungal	Helicobacter species DNA	Conventional culture, AFB culture	No	No	Helicobacter skin infection	Doxycycline, metronidazole, and amoxicillin-clavulanic acid started
10	Fungal forms seen on pathology following resection of metastatic osteosarcoma	No	Lung tissue	19	Bacterial, fungal	Acremonium species or Phialemonium species	Targeted serologies, PPD, galactomannan, 1,3 β-d-glucan, antigens for endemic fungi	No	Yes, numerous yeast forms are present on GMS stains of lung and lymph node	Acremonium or Phialemonium lung infection	Voriconazole started
11	Cardiac vegetation in patient s/p prosthetic valve placement	No	Cardiac hardware	1	Bacterial, fungal	Mycoplasma pneumoniae (detected with both broad-range and targeted PCR)	Conventional culture, targeted PCR, targeted serologies	Yes, Mycoplasma IgG and IgM	No	Mycoplasma pneumoniae endocarditis	Vancomycin and ceftriaxone discontinued, levofloxacin started
12	Incidentally identified abscess at time of scheduled cardiac homograft placement	No	Cardiac hardware	14	Bacterial	None	Conventional culture, next-generation sequencing	No	No	Sterile abscess believed to be foreign body reaction	Ceftriaxone and daptomycin discontinued
13	Fever, knee pain/swelling	Yes, active chemotherapy and s/p solid organ transplant	Bone biopsy	2	Bacterial	None	Conventional culture, AFB culture, targeted serologies, galactomannan, 1,3 β-d-glucan	No	No	Posttransplant lymphoproliferative disorder	Vancomycin and ceftriaxone discontinued
14	Fevers, liver lesions	Yes, on immunosuppression	Liver tissue	3	Bacterial, mycobacterial, fungal	None	Conventional culture, AFB culture, targeted PCR	Tissue culture growing Kocuria kristinae	No	EBV-induced lymphoma	Ampicillin discontinued
15	Fever	No	Pleural fluid	4	Bacterial, mycobacterial, fungal	Streptococcus pneumoniae	Conventional culture, AFB culture, targeted PCR, PPD, IGRA	No	Not sent	Streptococcus pneumoniae complicated pneumonia	Linezolid discontinued
16	Culture-negative sepsis	Yes, on immunosuppression	Lymphatic fluid	8	Bacterial, mycobacterial, fungal	None	Conventional culture, AFB culture, next-generation sequencing	No	Not sent	Lymphatic malformation without active infection	Meropenem discontinued
17	Lung mass	No	Lung tissue	16	Fungal	None	Conventional culture, galactomannan, 1,3 β-d-glucan	No	Yes, filamentous fungal forms, no definitive yeast forms	Possible fungal infection without definitive organism identified with significant toxicities from empiric therapy	Voriconazole discontinued
18	Endocarditis in patient s/p surgery for CCHD	No	Cardiac hardware	1	Bacterial, fungal	None	Conventional culture, AFB culture	Propionibacterium (Cutibacterium) acnes from blood culture	Yes, Gram-positive cocci	Infective endocarditis due to Gram-positive cocci	Ceftriaxone and rifampin discontinued
19	Fever, echocardiogram concerning for endocarditis	No	Native cardiac valve tissue	6	Bacterial	Streptococcus sanguinis	Conventional culture, AFB culture, targeted serologies	No	Yes, Gram-positive cocci	Infective endocarditis of native valve due to Gram-positive cocci	Vancomycin discontinued

^aAFB, acid-fast bacilli; BAL, bronchoalveolar lavage; CCHD, critical congenital heart disease; CMV, cytomegalovirus; DFA, direct fluorescent antibody; HSCT, hematopoietic stem cell transplant; IGRA, interferon gamma release assay; PCR, polymerase chain reaction; PJP, Pneumocystis jirovecii pneumonia; PPD, purified protein derivative; s/p, status post; SCID, severe combined immunodeficiency; UW, University of Washington.

Data on antimicrobial treatment in the 24 h prior to sample collection were available for 377/382 (99.0%) samples; 254 (64.7%) samples were collected from patients who had received antibiotics and/or antifungals in the preceding 24 h. Receipt of antimicrobials in this time frame was not significantly associated with either the rate of organism identification on BRPCR (P = 0.79) or whether BRPCR results changed clinical management (P = 0.62). Immunocompromised status was not significantly associated with identification of an organism on BRPCR (P = 0.80) or change in clinical management due to the result (P = 0.10). Pathology results were available for 313/382 (81.9%) samples. Sixty-seven (21.4%) pathology samples had findings consistent with a bacterial or fungal organism; the presence of an organism on pathology tissue staining was significantly associated with identification of an organism on BRPCR (P = 0.0002) but not with a change in clinical management (P = 0.54).

For samples with an organism identified, the median number of days between sample collection and receipt at the University of Washington was 5 days (IQR, 3 to 8); for those without an organism identified, it was also 5 days (IQR, 2 to 8.5). Data on time between collection and receipt were not available for 10 samples. Histograms displaying time between sample collection and receipt, stratified by identification of organism and by change to clinical management, are shown in Fig. 1.

FIG 1.

Days between sample collection and receipt at University of Washington Molecular Microbiology Library. (A) No organism identified. (B) Organism identified. (C) No change to clinical management. (D) Change to clinical management.

Results stratified by sample type are detailed in Table 2. The most frequently sent sample types were bronchoalveolar lavage (BAL) fluid (45), lung tissue (41), and bone (39). Cardiac hardware was the sample type with the greatest proportion of results that impacted clinical management (4/12, 33.3%). BAL samples had the greatest proportion of samples with positive BRPCR results that did not impact clinical management (16/45, 35.6%). There was no statistically significant difference among sample types in terms of the impact on clinical management of any BRPCR result (P = 0.06) or positive BRPCR result (P = 0.26).

No results from cerebrospinal fluid (29), intestinal tissue (21), wound tissue (7), joint fluid (6), bone marrow (6), synovial tissue (5), pericardial fluid (4), peritoneal fluid (4), or brain tissue (3) affected antimicrobial management.

DISCUSSION

To our knowledge, this report represents the largest analysis to date of BRPCR tests sent from a pediatric center. We found that BRPCR results led to a change in clinical management 5.0% of the time, with the rate of impact varying widely by sample type.

Given the lack of gold standard, we were not able to assess differences in sensitivity among sample types. However, we found that among all sample types, cardiac hardware had the highest rate of impacting clinical management. This testing is mentioned but not formally recommended in the American Heart Association endocarditis guidelines for both children (18) and adults (7), and a growing body of literature suggests the utility of BRPCR testing in culture-negative endocarditis in adults (8–11). In pediatric patients, the majority of available data about the use of BRPCR in endocarditis is limited to reports of individual cases (19–21). BRPCR does not offer an easy diagnostic strategy for endocarditis, as it is less sensitive on blood samples than tissue (12), so invasive tissue specimens are important for increasing the yield. This low sensitivity of BRPCR on blood samples parallels the reported low sensitivity of cell-free DNA, another molecular test sometimes performed on plasma samples in patients with culture-negative endocarditis (21–23). The poor yield of BRPCR on blood specimens was well-recognized within our institution, and no samples in our cohort were sent from blood.

When an appropriate sample is available, the identification of an organism can have significant impact on patient care, since the recommended antibiotic regimens for culture-negative endocarditis are very broad, and a positive result could permit narrowing of antibiotic spectrum (7, 18). Taken together with adult data, our findings suggest that, in the appropriate clinical setting, BRPCR could be a helpful tool in evaluation of pediatric culture-negative endocarditis.

In our cohort, positive results for over a third of BAL specimens were deemed to represent nonpathogens. Six of the 18 positive BAL samples had multiple organisms identified, and these were believed to represent upper airway flora and not true pathogens. Clinicians’ instinct to use BRPCR testing for this specimen type is understandable, as bronchoscopy is a less invasive method of sample collection than lung biopsy, yet rates of bacterial identification from BAL via conventional microbiological methods can be as low as 28% (24). Additionally, prior research has suggested that sending BRPCR from BAL specimens is useful when the sample is pretreated with antibiotics (25). Unfortunately, while BRPCR can increase the microbiologic yield from BAL fluid samples, interpreting these results is frequently challenging because the bronchoalveolar tree is not a sterile site. A prospective study comparing conventional bacterial culture and 16S bacterial sequencing sent from the same specimens found that BRPCR was able to identify organisms that were not cultured, but the authors acknowledged that the clinical significance of these organisms was not clear (24). In a prior cohort of pediatric patients, Lucas et al. found that although BAL specimens were the most common sample type, there were no situations in which an organism was detected from BAL fluid only by BRPCR and the result changed management (19).

A high rate of non-pathogen identification can lead to challenging antimicrobial management decisions for medical teams. In our cohort, as in other retrospective studies, it is difficult to evaluate how many patients ultimately received unnecessary treatment. This was epitomized in one case from our institution in which a team ultimately decided to treat Tropheryma whipplei identified on BRPCR from a BAL specimen despite expressing significant doubt in clinical notes that this organism represented a true pathogen (Table 3). These findings suggest that the decision to send BRPCR testing from BAL fluid should incorporate recognition of this limitation, and results should be interpreted cautiously, with providers recognizing the possibility that identified organisms are not causing disease.

Overall, our results were similar to those reported for a large predominantly adult cohort in which BRPCR affected clinical management approximately 4% of the time (6). The diagnostic yield of this test varies widely in the published literature, with a reported impact in 6% (19) and 25% (26) of cases for two smaller cohorts. These disparate values likely reflect wide differences in use patterns across institutions. This variation, along with concerns about rising health care costs and reports challenging the value of unrestricted use of BRPCR testing (2, 27, 28), underscores the need for additional data and guidelines to support best practices.

Anecdotally, a commonly cited reason for sending BRPCR testing from our institution is to make sure that the care team is not missing something, with the expectation that a negative test may provide some unmeasured reassurance. Through our retrospective chart review, we were not able to capture whether this benefit did in fact occur, but it is important to be cautious if using this test for its negative predictive value. Given the decreased sensitivity of BRPCR compared to targeted PCRs (4, 5), a negative BRPCR could provide false reassurance. In addition, as discussed above, identification of organisms of uncertain pathogenicity and clinical significance is not uncommon and could result in unnecessary treatment.

Our study has multiple limitations. BRPCR is frequently used at our institution, as elsewhere, in situations when a diagnosis is difficult to obtain. Therefore, no gold standard exists against which to calculate test performance characteristics, and we relied on the judgement of the treating providers regarding clinical significance of results. Using this approach has the potential to over- or underestimate the clinical utility of this test. Furthermore, we were not able to evaluate how many patients received unnecessary treatment in the setting of a BRPCR result that did not represent a true pathogen. Additionally, due to the retrospective nature of this study, we could not reliably collect certain data that would have been beneficial in understanding test performance. For example, whether tissue specimens were fresh or paraffin-embedded was not routinely recorded in the medical record; understanding performance characteristics of this test based on preservation status is an important question for future research. We were also unable to report test turnaround time, a variable that may have impacted clinical utility. Because this test is sent from our institution on a case-by-case basis and not in a protocolized fashion, our results with regard to effect on clinical management may not be generalizable to other settings. Other sites may find it helpful to track the impact of this test in their own setting to guide whether changes in local practice would be appropriate. Although we believe this is the largest report of samples sent from a pediatric center to date, the number of tests sent for some sample types was low, and our study therefore was likely underpowered to evaluate differences in clinical impact among sample types.

Prospective, multicenter studies on BRPCR could be helpful to reduce bias in interpretation of test results. Additionally, formal cost-effectiveness analyses would help in interpreting these data and guiding whether diagnostic stewardship interventions to change usage patterns would be appropriate. Overall, our findings demonstrate that only a limited number of BRPCR results at our institution led to a change in management, suggesting room for diagnostic stewardship and guideline development to optimize test usage.