Background
Plasma metagenomic next-generation sequencing (mNGS) is a hypothesis-generating test used in the diagnosis of infectious diseases (ID). While this technology has been commercially available since 2016, test performance characteristics and the optimal use cases are not yet well defined, Reference Niles, Lee, Lamb, Dhaheri and Boguniewicz1,Reference Blauwkamp, Thair and Rosen2 and use varies widely between centers. Reference Naureckas Li, Blumenthal and Sick-Samuels3
At our hospital, we noted plasma mNGS use disproportionately higher than peer institutions. Reference Naureckas Li, Blumenthal and Sick-Samuels3 To support appropriate use, we instituted a second-sign process that required ID approval before the sample could be collected. Reference Naureckas Li, Jordan, Haymond, Koscinski and Jhaveri4 Despite moderate reductions in usage, we recognized that our division had significant provider variation in determination of appropriate use of plasma mNGS testing, which threatened the acceptance of the second sign procedure by other medical teams. We therefore undertook a modified Delphi consensus to achieve local alignment on situations in which this test would be appropriate. Reference FItch, Bernstein and Aguilar5
Methods
We completed this project at our freestanding pediatric hospital with approximately 350 beds in the upper Midwest. At the time of this work, our hospital’s division of ID had 17 attending physicians who saw patients on the inpatient service, three nurse practitioners, and three active clinical fellows. Our division has two consult-only services: one that sees patients pre- and post-transplant and another that consults on all other patient types, including those who are immunocompromised for other reasons. Four attendings worked only on the transplant team, six only on the general team, and the remainder on both. All attendings cover both teams on weekend call, and all were invited to participate in the modified scenario Delphi consensus process. Reference Mukherjee, Hugé and Sutherland6
We planned three consensus rounds using a modified structure in that the first and second rounds were anonymous online voting, but the third round was an in-person open discussion with anonymous voting. We prospectively set a threshold of ≥75% agreement as consensus. For Round 1, 19 clinical scenarios (Table 1) were developed by two attending ID physicians (CNL and RJ) to span the most frequent clinical situations for mNGS requests on both the general and transplant services. A survey containing these scenarios was sent to all providers in the division via REDCap (Research Electronic Data Capture, Vanderbilt, TN). Providers were asked to answer whether they would approve plasma mNGS testing for each scenario, answer choices were yes/no/depends on context. If providers selected “depends on context,” they were asked to provide free-text clarification.
Table 1. Summary of Modified Delphi Consensus Results
† For votes totaling less than 100%, remaining answers were “depends on context”
*Skipped for Round 3, presumed NO since consensus for 7 days is NO.
Abbreviations: ICU – intensive care unit, ID – infectious diseases, mNGS – metagenomic next-generation sequencing, NOS – not otherwise specified, VAD – ventricular assist device.
Following Round 1, scenarios with significant disagreement (in which there were ≥ 4 “yes” and ≥ 4 “no” responses) were held for the Round 3 in-person deliberation. For the questions that had not achieved consensus but were not split significantly, wording was refined by CNL and RJ using the qualitative responses to the “it depends” selections. These modified questions were then sent via REDCap survey for Round 2, again with answer choices of “yes/no/depends on context.”
Round 3 took place in-person with a virtual option for those who were unable to be on site, which was the standard format for all clinical conferences and faculty meetings within our division. The meeting was recorded for later analysis.
Data were analyzed using descriptive statistics in MS Excel. This work was deemed exempt from full board review by the Lurie Institutional Review Board.
Results
Nineteen (83%) providers participated in the first round of the modified Delphi process, 20 (87%) in the second, and 14 (61%) in the third. In the Round 3, 13 participated in-person and one was present virtually.
Following Round 1, three scenarios reached consensus; all three against use of plasma mNGS. In Round 2, two additional scenarios reached consensus for using plasma mNGS, and six additional reached consensus for not using plasma mNGS testing. By the end of Round 3, our division ultimately reached consensus for a total of 17 clinical scenarios: 3 consensus to use plasma mNGS and 14 consensus to not use plasma mNGS. Four scenarios did not reach consensus (Table 1). Two scenarios were each split into two separate scenarios during the refinement process. The full text of the questions and results are shown in Supplemental Figure 1.
All four scenarios that did not reach consensus were scenarios involving immunocompromised patients. The only scenario in which the group supported use of plasma mNGS in immunocompetent patients was in those cases with high clinical suspicion for endocarditis when cultures were negative and the patient was not improving on empiric therapy (Table 1).
Discussion
Through a modified Delphi process, our pediatric ID division was able to reach consensus on appropriate use of plasma mNGS in the complex patient population we serve for 17 different scenarios for both immunocompetent and immunocompromised patients.
In a recent survey of ID providers and microbiology laboratory directors, nearly 2/3 believed that full ID consultation should be required before plasma mNGS is sent. However, the same survey found wide variation in perceptions of which clinical scenarios represented appropriate use of this test. Reference Naureckas Li, Blumenthal and Sick-Samuels3 While there are some data on the use of utilization committees that focus on cost-effective test use for approval of plasma mNGS, Reference Shean, Garrett, Malleis, Lieberman and Bradley7,Reference Goggin, Fox, Jaggi, Shane and Gonzalez8 there are limited outcomes data from sites that require ID provider consultation or approval, and there is minimal guidance available to support these providers in making decisions around whether to approve testing. In the absence of strong data and national or international guidelines to support these choices, consensus recommendations can serve as a stopgap measure to fill this void.
We were not surprised that areas in which our providers had the greatest consensus were those that had the most support in the literature. The use of plasma mNGS was recently added to international endocarditis guidelines, Reference Fowler, Durack and Selton-Suty9 which was the likely driver of our universal consensus. Similarly, although prospective data are limited, a multicenter prospective study concluded that the use of plasma mNGS in addition to standard of care testing increased the rate of organism detection in patients with pneumonia significant immunocompromise. Reference Bergin, Chemaly and Dadwal10 These data likely drove the high level of consensus in favor of using this test for suspected lung infections in immunocompromised patients when bronchoscopy could not be performed.
These data have significant limitations. Due to conflicting obligations, not all providers participated in each round. As data collection was anonymous, we are not able to fully understand the degree of selection bias this may have introduced. These findings come from a single center, and with the lack of high-quality data, provider opinions are likely to be influenced by local culture. However, in the absence of national or international guidance on appropriate test utilization, we hope these results can be used to support diagnostic stewardship efforts in sites without access to pediatric ID expertise and serve as a framework for other referral centers to reach consensus about use of this test based on their local environment.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2025.10334
Acknowledgements
The authors have no conflicts of interest or financial support relevant to this work to disclose. They would like to thank the members of the Lurie Children’s Hospital Division of Infectious Diseases for participating in this modified Delphi consensus process.
Financial support
There was no specific financial support for this work
Competing interests
There are no conflicts of interest to report.
Open access
