Discussion
Our study prospectively evaluated the utility of FebriDx to rapidly identify suspected cases of COVID-19 disease in a hospital / ED setting in an effort to improve isolation and cohorting procedures. Forty-nine patients were enrolled, and one patient was excluded after enrollment due to the inability to obtain enough blood for FebriDx testing after the first attempt. Due to the patient’s advanced age and unstable clinical state at the time of testing, FebriDx testing was not repeated and thus this patient was excluded from the analyses. Therefore, 48 patients were included in the analyses of the for the primary and secondary endpoints. After using PHE screening criteria for suspected COVID-19 signs/symptoms, FebriDx correctly identified all patients (35/35) who met the Case Definition30,31 for COVID-19 (Figure 3). FebriDx also correctly identified all patients that had a bacterial infection by case definition (8/8) which resulted in a diagnostic sensitivity of 100%. FebriDx had three false positives for bacterial infection in two non-infectious patients and one clinically indeterminate patient demonstrating a specificity for bacterial infection of 92.5% (37/40) (Table 2). T he mortality rate for COVID-19 related deaths was 33.3% our cohort. This is in keeping with COVID-19 related mortality rates for hospitalised patients within our health system as well as the national mortality figures for hospitalised patients.35 Standalone conventional CRP was elevated in both COVID-19 viral cases and bacterial cases in our study (median [IQR] 76 mg/L [50.3-115.5]; 94 mg/L [59.5.0-152.0], respectively). Brendish et al., also found CRP to be elevated in both COVID and Non-COVID-19 cases (median (range) 83 mg/L (32-136 mg/L); 33 (9-114 mg/L) respectively p<0.0001).36Although Brendish et al. found the difference to be statistically significant, the considerable overlap of quantitative CRP may make it difficult to differentiate viral from bacterial infection as a standalone test.36 The same appears to apply to procalcitonin and leukocyte count in our study (Table 1). MxA confers the diagnostic sensitivity and specificity needed to differentiate elevated CRP associated with viral vs. bacterial infection and may help to avoid mixing non-COVID-19 with COVID-19 whilst awaiting the results of rRT-PCR that can take up to 48 hours in the hospital/ED settings. In our study 7 patients, who in the end were not diagnosed with COVID-19, were inadvertently exposed to COVID-19 due to the unintended mixed cohorting that occurred whilst awaiting swab rRT-PCR results. As our study was intended to evaluate diagnostic accuracy of FebriDx as part of an initial triage strategy, the FebriDx test results were not used to make decisions regarding cohorting until after the study was concluded and the results were analysed. Based on the high NPV of FebriDx in our setting, it is possible that the unintended exposure of non-COVID-19 patients could have been avoided if FebriDx was utilised part of the initial triage of ARI patients with suspected COVID-19. Utilising FebriDx for enabling cohorting decisions could have avoided exposure in these cases.
A recent study by Brendish and colleagues from University Hospital Southampton, Southampton, England, also evaluated the diagnostic accuracy of FebriDx in 248 hospitalised adults who presented with suspected COVID-19 regardless of duration of symptom onset (inclusive of the PHE Case Definition for Possible Infection).36,37Of the 248 patients who underwent FebriDx and SARS-CoV-2 rRT-PCR, 118 had SARS-CoV-2 detected (prevalence 48%). Diagnostic sensitivity, specificity, NPV and PPV were 93%, 86%, 86% and 93%, respectively. Despite some methodological differences their results were comparable to our study.
Based on the diagnostic performance characteristics of FebriDx demonstrated in our study as well as Brendish et al.36, we propose that in the current SARS-CoV-2 pandemic situation, patients presenting with signs and symptoms of ARI and suspected of COVID-19 infection should be tested with FebriDx test as part of the initial diagnostic triage process. Those testing ‘viral positive’ (+MxA), should be treated as ‘positive COVID-19’ and cohorted with other COVID-19 positive patients. This would help avoid unnecessary exposure to other suspected patients who may turn out to be negative on confirmatory rRT-PCR testing. If FebriDx result is ‘viral negative’ an alternative diagnosis such as bacterial infection or non-infectious conditions such as bacterial pneumonia or LRTI, should be considered at the outset. It should be noted however, that patient enrollment took place at the peak of the COVID-19 outbreak in our region and pre-test probability of recruiting COVID-19 positive patients was relatively high. This strategy allowed us to obtain a maximum number of potential COVID-19 infections to evaluate FebriDx-based identification of SARS-CoV-2 infection. Therefore, the prevalence of COVID-19 infection was 73.2% in our setting and this may have increased the chance of obtaining a high PPV.
Future viral outbreaks and seasonal infections could be managed, ideally, by optimizing all available diagnostic tools (e.g. clinical assessment, host response, molecular testing, antibody testing etc.).13 Pulia et al. proposed ‘Multi-tiered Screening and Diagnostic Strategy’ that incorporates a comprehensive approach that could be used in the SARS-CoV-2 pandemic and potentially as a general strategy in future pandemics.13 The strategy proposes that after initial screening (e.g. clinical signs/symptoms of the suspected infection), such as the initial screening performed in our study, patients could be (i) be quickly tested for a viral, bacterial or absent immune response to an infection, followed by (ii) rapid confirmatory pathogen-specific testing; and (iii) rapid antibody testing could be performed in patients that present with greater than 7 days of symptom onset to confirm a recent or past infection. Although FebriDx should not be used as a surrogate for pathogen-detection tests, it can be applied to rapidly categorise patients as having bacterial or viral infections or non-infectious conditions as part of the diagnostic triage process.36 This would allow bacterial infections/non-infectious conditions to be cohorted separately from suspected viral infections. Those with viral infections would go on to have confirmatory testing to improve cohorting within the viral category, whereas antibiotics could be considered for patients positive for bacterial infection. Repeat rRT-PCR testing could be considered in high risk patients who test viral positive on FebriDx but have a negative initial SARS-CoV-2 PCR.
Our study is not without limitations. Based on the urgent need to improve testing turnaround times and patient isolation strategies at our hospital, it was not possible to design and perform a multi-centre trial that included a control group. Antibody testing was not available for all patients enrolled nor is antibody testing required by PHE, ECDC nor CDC for confirmation of COVID-19 infection. That said, antibody confirmation would be ideal for determining definitive COVID-19 infection after 14 days of symptom onset, especially in cases that have a high clinical suspicion but were SARS-CoV-2 is not detected by rRT-PCR testing. Due to the lack of a gold standard test for COVID-19 infection, the current assessment of the performance characteristics of rRT-PCR may suffer from an incorporation bias. We attempted to mitigate this by including clinical, radiological, and epidemiological criteria for final diagnosis of COVID-19 infection as is consistent with the ECDC and CDC Case Definitions. Finally, patients presenting in our hospital with COVID-19 symptoms were generally adults. Therefore, additional studies would be required to assess this strategy in children.
At the moment, in our clinical setting, and according to overwhelming data reports by the PHE, CDC and ECDC, the predominant virus causing hospitalisation amongst adults at present, seems to be SARS-CoV-2.28 Based on our study findings, we provide evidence that FebriDx could be deployed as part of the initial diagnostic triage process for early identification of symptomatic COVID-19 patients presenting in a hospital setting.
Disclosures
Authors declare that they have no conflicts of interest.