Key Points
Question
What is the ability of a host gene expression test to accurately discriminate bacterial from viral infection among patients with acute respiratory illness?
Findings
In this diagnostic study involving analysis of 616 children and adults with febrile acute respiratory illness of 7 or fewer days’ duration, the host response bacterial/viral test had up to 90% sensitivity, 82% specificity, and 98% negative predictive value for bacterial infection, which was significantly better than procalcitonin measurement.
Meaning
The study’s findings suggest that an accurate point-of-need host response test with high negative predictive value may identify patients unlikely to have bacterial infection, offering a better antibiotic stewardship strategy than is currently available.
Abstract
Importance
Bacterial and viral causes of acute respiratory illness (ARI) are difficult to clinically distinguish, resulting in the inappropriate use of antibacterial therapy. The use of a host gene expression–based test that is able to discriminate bacterial from viral infection in less than 1 hour may improve care and antimicrobial stewardship.
Objective
To validate the host response bacterial/viral (HR-B/V) test and assess its ability to accurately differentiate bacterial from viral infection among patients with ARI.
Design, Setting, and Participants
This prospective multicenter diagnostic study enrolled 755 children and adults with febrile ARI of 7 or fewer days’ duration from 10 US emergency departments. Participants were enrolled from October 3, 2014, to September 1, 2019, followed by additional enrollment of patients with COVID-19 from March 20 to December 3, 2020. Clinical adjudication of enrolled participants identified 616 individuals as having bacterial or viral infection. The primary analysis cohort included 334 participants with high-confidence reference adjudications (based on adjudicator concordance and the presence of an identified pathogen confirmed by microbiological testing). A secondary analysis of the entire cohort of 616 participants included cases with low-confidence reference adjudications (based on adjudicator discordance or the absence of an identified pathogen in microbiological testing). Thirty-three participants with COVID-19 were included post hoc.
Interventions
The HR-B/V test quantified the expression of 45 host messenger RNAs in approximately 45 minutes to derive a probability of bacterial infection.
Main Outcomes and Measures
Performance characteristics for the HR-B/V test compared with clinical adjudication were reported as either bacterial or viral infection or categorized into 4 likelihood groups (viral very likely [probability score <0.19], viral likely [probability score of 0.19-0.40], bacterial likely [probability score of 0.41-0.73], and bacterial very likely [probability score >0.73]) and compared with procalcitonin measurement.
Results
Among 755 enrolled participants, the median age was 26 years (IQR, 16-52 years); 360 participants (47.7%) were female, and 395 (52.3%) were male. A total of 13 participants (1.7%) were American Indian, 13 (1.7%) were Asian, 368 (48.7%) were Black, 131 (17.4%) were Hispanic, 3 (0.4%) were Native Hawaiian or Pacific Islander, 297 (39.3%) were White, and 60 (7.9%) were of unspecified race and/or ethnicity. In the primary analysis involving 334 participants, the HR-B/V test had sensitivity of 89.8% (95% CI, 77.8%-96.2%), specificity of 82.1% (95% CI, 77.4%-86.6%), and a negative predictive value (NPV) of 97.9% (95% CI, 95.3%-99.1%) for bacterial infection. In comparison, the sensitivity of procalcitonin measurement was 28.6% (95% CI, 16.2%-40.9%; P < .001), the specificity was 87.0% (95% CI, 82.7%-90.7%; P = .006), and the NPV was 87.6% (95% CI, 85.5%-89.5%; P < .001). When stratified into likelihood groups, the HR-B/V test had an NPV of 98.9% (95% CI, 96.1%-100%) for bacterial infection in the viral very likely group and a positive predictive value of 63.4% (95% CI, 47.2%-77.9%) for bacterial infection in the bacterial very likely group. The HR-B/V test correctly identified 30 of 33 participants (90.9%) with acute COVID-19 as having a viral infection.
Conclusions and Relevance
In this study, the HR-B/V test accurately discriminated bacterial from viral infection among patients with febrile ARI and was superior to procalcitonin measurement. The findings suggest that an accurate point-of-need host response test with high NPV may offer an opportunity to improve antibiotic stewardship and patient outcomes.
This diagnostic study assesses the ability of a host gene expression test to accurately discriminate between bacterial and viral infection among children and adults with acute febrile respiratory illness.
Introduction
Acute respiratory illness (ARI) is the most common reason for urgent health care visits.1,2 Routinely available clinical information inadequately differentiates infections with bacterial causes from those with viral causes, contributing to high rates of inappropriate antibiotic medication use.3,4,5,6,7 Tests that reliably discriminate bacterial from viral infections could decrease diagnostic uncertainty, reduce inappropriate use of antibacterial therapy, and improve patient outcomes.
Pathogen identification tests, such as multiplexed syndromic panels, are important diagnostic tools but are unable to detect a bacterial or viral cause in most ARI cases.8 They also do not distinguish colonization from infection when a microbe is identified.9,10 Because the immunological responses to bacterial and viral infection are distinct, measuring the host response overcomes these limitations. Procalcitonin, the most widely used host biomarker, has exhibited mixed results for bacterial vs viral discrimination11,12,13 and for guiding the use of antibacterial therapy.14,15 Another approach involves measuring peripheral blood host gene expression, which can now be performed using clinically available platforms at the point of need.16,17
A previous study18 described the discovery of a gene expression signature discriminating bacterial from viral illness. This signature was further developed into a research-use-only host response bacterial/viral (HR-B/V) test via the widely used BioFire system and was evaluated using banked samples primarily from adults in a geographically limited environment.16 The present diagnostic study aimed to validate this HR-B/V test among a multisite diverse prospectively enrolled cohort and assess the superiority of the HR-B/V test over procalcitonin measurement. The study examined the feasibility and potential utility of a rapid point-of-need host response test to differentiate bacterial from viral respiratory infections.
Methods
Study Design
This study was approved by the institutional review boards of each participating hospital. All participants or legally authorized representatives provided written informed consent. The study followed the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline for diagnostic studies.19
Participants were prospectively enrolled using convenience sampling from 10 US emergency departments: Duke University Hospital (Durham, North Carolina), Durham VA Health Care System (Durham, North Carolina), UNC Health Care (Chapel Hill, North Carolina), Henry Ford Health System (Detroit, Michigan), Brigham and Women’s Hospital (Boston, Massachusetts), University of California Hospital at Davis (Sacramento); Children’s Memorial Hermann (McGovern Medical, Houston, Texas), Hasbro Children’s Hospital (Providence, Rhode Island), University of Utah Medical Center (Salt Lake City), Children’s Hospital of Pittsburgh (Pittsburgh, Pennsylvania) and Newton-Wellesley Hospital (Newton, Massachusetts). Participants were enrolled from October 3, 2014, to September 1, 2019, followed by additional enrollment of patients with COVID-19 from March 20 to December 3, 2020. Included participants were 2 years or older with febrile ARI of 7 or fewer days’ duration. Acute respiratory illness was defined as having 2 or more qualifying symptoms or 1 qualifying symptom and at least 1 qualifying vital sign abnormality. A list of qualifying symptoms is available in eMethods in Supplement 1. Age was categorized as follows: children (2-11 years), adolescents (12-21 years), adults (22-64 years), and older adults (≥65 years). Patients with a known or suspected infection at any other anatomic site requiring antibacterial therapy were excluded. Charlson Comorbidity Index scores were calculated as previously reported.20
Enrollment ended before the onset of the COVID-19 pandemic. We therefore supplemented this analysis post hoc by analyzing blood RNA samples (PAXgene Blood RNA Tubes; QIAGEN) collected from 33 participants with acute SARS-CoV-2 infection through convenience sampling at Duke University or in a community setting from March to December 2020. All participants had positive results for SARS-CoV-2 on a polymerase chain reaction test, negative results for immunoglobin G, 7 or fewer days of symptoms, and no known bacterial coinfection.
Reference Standard
Two adjudicators (including E.R.K., N.A., J.G., G.H., J.P., N.J., I.U., S.A., L. Mercurio, T.H.C., L. May, R.W.H., J.E.L., S.H.G., D.J.P., D.S.H., K.A., R.J., L.G.T., M.T.M., C.W.W., and E.L.T. along with other adjudicators who were not authors) from each enrollment site were randomly assigned to independently assess the likelihood of bacterial infection, viral infection, or no infection. Details regarding the adjudication process can be found in eMethods in Supplement 1. The primary analysis included participants with a high-confidence reference adjudication, which was defined as adjudicator concordance and the presence of an identified pathogen confirmed by microbiological testing (Figure 1; eTable 1 in Supplement 1). The secondary analysis involved all participants with a bacterial or viral infection, including those with a low-confidence adjudication, which was defined as adjudicator discordance or the absence of an identified pathogen in microbiological testing (Figure 1; eTable 1 in Supplement 1).
Figure 1. Study Flowchart.
aThe 33 patients in the COVID-19 cohort are an entirely separate group, enrolled independently of the rest of the cohort, which is why they are intentionally depicted in parallel to the other analysis groups.
Host Response Bacterial vs Viral Testing
The HR-B/V test detects 45 host messenger RNA targets using real-time quantitative reverse transcription polymerase chain reaction testing in approximately 45 minutes; the test was developed in collaboration with BioFire Diagnostics.18 RNA-preserved blood (PAXgene Blood RNA Tubes; QIAGEN) was processed according to manufacturer instructions and stored at −80 °C. The HR-B/V test was later performed by loading 100 μL of preserved blood (approximately 27 μL whole-blood volume) directly into research-use-only HR-B/V pouches (BioFire Diagnostics), which measured the relative abundance of target messenger RNAs normalized to the expression of 3 housekeeping genes (DECR1, PPIB, and TRAP1). Testing was conducted at Duke University. Details regarding the test and genes included in the panel have been previously published.16
Statistical Analysis
Test measurements for the primary and secondary analysis included sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood ratios, and area under the receiver operating characteristic curve.21 The HR-B/V test results were reported as probabilities of bacterial infection using a single threshold (probability score >0.41). We also grouped results into 4 likelihood categories: viral very likely (probability score <0.19), viral likely (probability score of 0.19-0.40), bacterial likely (probability score of 0.41-0.73), and bacterial very likely (probability score >0.73). The thresholds for these interpretive groups were derived from a previously characterized cohort.16
A procalcitonin level of 0.25 ng/mL or higher indicated bacterial infection. We also used previously described likelihood groups (<0.10 ng/mL indicated bacterial very unlikely, 0.10-0.24 ng/mL indicated bacterial unlikely, 0.25-0.50 ng/mL indicated bacterial likely, and >0.50 ng/mL indicated bacterial very likely).22 The median value was imputed for 25 participants with missing data.
We estimated the impact of age, sex, race and ethnicity, geographic area, hospital admission status, comorbidities, previous antibiotic medication use, illness duration, and illness severity for test performance using analysis of variance. Race and ethnicity were participant-defined or based on medical records when participants could not answer. Race was included as a covariate to determine whether it impacted the test, which measures immunological responses to infection. Bias-corrected and accelerated bootstrapping was used to estimate 95% CIs, and permutation tests were used to assess the statistical significance of the differences between performance measurements.23 All analyses were performed using Matlab software, version 2016b (Mathworks). Values for the area under the receiver operating characteristic curve were compared using the DeLong method.24 The significance threshold was 2-tailed P = .05.
Results
Participant Characteristics
Among 755 participants enrolled in the study, the median age was 26 years (IQR, 16-52 years); 360 participants (47.7%) were female, and 395 (52.3%) were male. Participants were racially and ethnically diverse, with 13 individuals (1.7%) identifying as American Indian, 13 (1.7%) as Asian, 368 (48.7%) as Black, 131 (17.4%) as Hispanic, 3 (0.4%) as Native Hawaiian or Pacific Islander, 297 (39.3%) as White, and 60 (7.9%) as unspecified (Table 1). Participants without an adjudicated bacterial or viral infection were excluded (Figure 1), resulting in 616 participants (median age, 24 years [IQR, 14-49 years]; 292 [47.4%] female and 324 [52.6%] male) included in the analyses. The median procalcitonin level among those participants was 0.05 ng/mL (IQR, 0-0.13 ng/mL) in patients with viral infection and 0.12 ng/mL (IQR, 0-0.47 ng/mL) in patients with bacterial infection (P = .004).
Table 1. Participant Demographic and Clinical Characteristics.
| Characteristic | Participants, No. (%) | |||||
|---|---|---|---|---|---|---|
| Enrolled | Included in analyses | Primary analysis | Secondary analysis | |||
| Bacterial infection | Viral infection | Bacterial infection | Viral infection | |||
| Total, No. | 755 | 616 | 49 | 285 | 117 | 499 |
| Demographic | ||||||
| Age, median (IQR), y | 26 (16-52) | 24 (14-49) | 24 (16-36) | 23 (10-48) | 29 (19-53) | 24 (12-48) |
| Sex | ||||||
| Female | 360 (47.7) | 292 (47.4) | 24 (49.0) | 139 (48.8) | 55 (47.0) | 237 (47.5) |
| Male | 395 (52.3) | 324 (52.6) | 25 (51.0) | 146 (51.2) | 62 (53.0) | 262 (52.5) |
| Race and ethnicity | ||||||
| American Indian | 13 (1.7) | 11 (1.8) | 0 | 5 (1.8) | 2 (1.7) | 9 (1.8) |
| Asian | 13 (1.7) | 11 (1.8) | 0 | 5 (1.8) | 1 (0.9) | 10 (2.0) |
| Black | 368 (48.7) | 293 (47.6) | 23 (46.9) | 135 (47.4) | 46 (39.3) | 247 (49.5) |
| Hispanic | 131 (17.4) | 110 (17.9) | 8 (16.3) | 58 (20.4) | 19 (16.2) | 91 (18.2) |
| Native Hawaiian or Pacific Islander | 3 (0.4) | 2 (0.3) | 0 | 2 (0.7) | 0 | 2 (0.4) |
| White | 297 (39.3) | 245 (39.8) | 20 (40.8) | 115 (40.4) | 56 (47.9) | 189 (37.9) |
| Unspecified | 60 (7.9) | 53 (8.6) | 6 (12.2) | 28 (9.8) | 12 (10.3) | 41 (8.2) |
| Clinical | ||||||
| Symptom duration, median (IQR), d | 3 (2-4) | 3 (2-4) | 3 (2-4) | 3 (2-4) | 3 (2-4) | 3 (2-4) |
| Procalcitonin level, median (IQR), ng/mL | 0.13 (0.08-0.32) | 0.07 (0-0.16) | 0.12 (0-0.47) | 0.06 (0-0.15) | 0.12 (0-0.47) | 0.05 (0-0.13) |
| Charlson Comorbidity Index score, mean (SD) | 1.37 (2.11) | 1.19 (1.94) | 1.35 (2.65) | 1.04 (1.60) | 1.74 (2.72) | 1.06 (1.68) |
| ICU admission | 35 (4.6) | 23 (3.7) | 2 (4.1) | 12 (4.2) | 5 (4.3) | 18 (3.6) |
| Deaths | 7 (0.9) | 4 (0.6) | 1 (2.0) | 0 | 4 (3.4) | 0 |
| Receipt of antibacterial therapy before enrollment | 103 (13.6) | 87 (14.1) | 17 (34.7) | 26 (9.1) | 43 (36.8) | 44 (8.8) |
Abbreviation: ICU, intensive care unit.
Host Response Bacterial/Viral Test Performance
Primary Analysis
The primary analysis included 334 participants (285 with viral infection and 49 with bacterial infection) with a high-confidence adjudication (eTable 1 in Supplement 1). Participants were first analyzed using a single threshold (probability score ≥0.41, indicating bacterial infection) (Figure 2A). In the primary analysis cohort, the HR-B/V test had sensitivity of 89.8% (95% CI, 77.8%-96.2%), specificity of 82.1% (95% CI, 77.4%-86.6%), and an NPV of 97.9% (95% CI, 95.3%-99.1%) for bacterial infection (Table 2; Figure 2B). Using a procalcitonin threshold of 0.25 ng/mL for bacterial infection, sensitivity was lower at 28.6% (95% CI, 16.2%-40.9%; P < .001) but specificity was higher at 87.0% (95% CI, 82.7%-90.7%; P = .006) compared with the HR-B/V test (Table 2; Figure 2C); the NPV was 87.6% (95% CI, 85.5%-89.5%; P < .001). The procalcitonin area under the curve was 0.59 (95% CI, 0.49-0.68) compared with 0.92 (95% CI, 0.87-0.94) for the HR-B/V test (P < .001) (eFigure in Supplement 1).
Figure 2. Test Performance.
A, Four-group likelihood schemes for the host response bacterial/viral (HR-B/V) test and procalcitonin measurement. In both cases (HR-B/V test vs procalcitonin measurement), results can also be interpreted in a dichotomous manner using a single threshold. B, The primary analysis cohort included 334 participants with high-confidence adjudications. The line within the box indicates the median value, the box indicates the first to third IQRs, the whiskers indicate the IQRs multiplied by 1.5, and the plus signs indicate outliers. C, The primary analysis cohort included 334 participants with high-confidence adjudications. The line within the box indicates the median value, the box indicates the first to third IQRs, the whiskers indicate the IQRs multiplied by 1.5, and the plus signs indicate outliers. D, The secondary analysis cohort included 616 participants with high- and low-confidence adjudications. The line within the box indicates the median value, the box indicates the first to third IQRs, the whiskers indicate the IQRs multiplied by 1.5, and the plus signs represent outliers. E, The secondary analysis cohort included 616 participants with high- and low-confidence adjudications. The line within the box indicates the median value, the box indicates the first to third IQRs, the whiskers indicate the IQRs multiplied by 1.5, and the plus signs indicate outliers. F, The COVID-19 cohort included 33 participants with acute SARS-CoV-2 infection. The dots in panels B to F indicate individual participants. BL indicates bacterial likely; BU, bacterial unlikely; BVL, bacterial very likely; BVU, bacterial very unlikely; VL, viral likely; and VVL, viral very likely.
Table 2. Performance Characteristics of HR-B/V Test vs Procalcitonin Measurement for the Diagnosis of Bacterial Infection Using a Single Threshold.
| Analysis | % (95% CI) | LR+ (95% CI) | LR− (95% CI) | AUROC (95% CI) | |||
|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | PPV | NPV | ||||
| Primary analysis | |||||||
| HR-B/V test | 89.8 (77.8-96.2) | 82.1 (77.4-86.6) | 46.3 (39.8-53.0) | 97.9 (95.3-99.1) | 5.02 (3.85-6.55) | 0.12 (0.05-0.29) | 0.92 (0.87-0.94) |
| Procalcitonin measurement | 28.6 (16.2-40.9) | 87.0 (82.7-90.7) | 27.5 (18.1-39.3) | 87.6 (85.5-89.5) | 2.20 (1.29-3.76) | 0.82 (0.68-0.99) | 0.59 (0.49-0.68) |
| Secondary analysis | |||||||
| HR-B/V test | 86.4 (79.6-92.5) | 71.9 (67.7-75.9) | 41.9 (38.1-45.8) | 95.7 (93.4-97.3) | 3.08 (2.63-3.60) | 0.19 (0.12-0.30) | 0.85 (0.81-0.89) |
| Procalcitonin measurement | 31.6 (23.3-39.5) | 87.3 (84.5-90.1) | 37.0 (29.2-45.5) | 84.5 (82.8-86.1) | 2.50 (1.76-3.56) | 0.78 (0.69-0.89) | 0.62 (0.56-0.67) |
Abbreviations: AUROC, area under the receiver operating characteristic curve; HR-B/V, host response bacterial/viral; LR−, likelihood ratio negative; LR+, likelihood ratio positive; NPV, negative predictive value; PPV, positive predictive value.
Dichotomizing results as either bacterial or viral with a single cutoff value (probability score ≥0.41) would not have accounted for the diagnostic confidence afforded by results in the highest or lowest groups. We therefore analyzed results using the 4-tier likelihood method. A total of 26 bacterial infections (53.1%) were classified as bacterial very likely (probability of >0.73), and 172 viral infections (60.4%) were classified as viral very likely (probability of <0.19) (Table 3; Figure 2B). In the viral very likely group, the NPV for bacterial infection was 98.9% (95% CI, 96.1%-100%). In the bacterial very likely group, the PPV for bacterial infection was 63.4% (95% CI, 47.2%-77.9%) owing to the high prevalence of viral infection (85.3%).
Table 3. Performance Characteristics of HR-B/V Test vs Procalcitonin Measurement for the Diagnosis of Bacterial Infection Using a 4-Group Likelihood Scheme.
| Test and group | No./total No. (%) | % (95% CI) | Interval likelihood ratio (95% CI)c | ||||
|---|---|---|---|---|---|---|---|
| Bacterial infection | Viral infection | PPV in groupa | NPV in groupa | Sensitivityb | Specificityb | ||
| Primary analysis | |||||||
| HR-B/V test | |||||||
| VVL (probability score <0.19) | 2/49 (4.1) | 172/285 (60.4) | NA | 98.9 (96.1-100) | NA | 60.4 (54.7-66.0) | 0.07 (0-0.20) |
| VL (probability score 0.19-0.40) | 3/49 (6.1) | 62/285 (21.8) | NA | 95.4 (88.1-98.7) | NA | 21.8 (17.0-26.9) | 0.28 (0-0.77) |
| BL (probability score 0.41-0.73) | 18/49 (36.7) | 36/285 (12.6) | 33.3 (22.9-48.3) | NA | 36.7 (23.4-50.8) | NA | 2.91 (1.76-4.74) |
| BVL (probability score >0.73) | 26/49 (53.1) | 15/285 (5.3) | 63.4 (47.2-77.9) | NA | 53.1 (38.8-67.4) | NA | 10.08 (5.69-17.36) |
| Procalcitonin measurement | |||||||
| BVU (<0.10 ng/mL) | 22/49 (44.9) | 187/285 (65.6) | NA | 89.5 (84.6-93.1) | NA | 65.6 (59.9-70.6) | 0.68 (0.49-0.94) |
| BU (0.10-0.24 ng/mL) | 13/49 (26.5) | 61/285 (21.4) | NA | 82.4 (72.0-90.4) | NA | 21.4 (17.1-26.8) | 1.24 (0.72-2.12) |
| BL (0.25-0.50 ng/mL) | 4/49 (8.2) | 13/285 (4.6) | 23.5 (6.7-50.0) | NA | 8.2 (2.3-18.3) | NA | 1.79 (0.41-5.82) |
| BVL (>0.50 ng/mL) | 10/49 (20.4) | 24/285 (8.4) | 29.4 (15.9-45.6) | NA | 20.4 (10.3-33.9) | NA | 2.42 (1.17-4.90) |
| Secondary analysis | |||||||
| HR-B/V test | |||||||
| VVL (probability score <0.19) | 8/117 (6.8) | 238/499 (47.7) | NA | 96.7 (94.2-98.4) | NA | 47.7 (43.6-52.3) | 0.14 (0.07-0.27) |
| VL (probability score 0.19-0.40) | 8/117 (6.8) | 121/499 (24.2) | NA | 93.8 (87.9-96.9) | NA | 24.2 (20.7-28.5) | 0.28 (0.12-0.56) |
| BL (probability score 0.41-0.73) | 36/117 (30.8) | 95/499 (19.0) | 27.5 (20.0-35.7) | NA | 30.8 (22.6-39.7) | NA | 1.62 (1.19-2.27) |
| BVL (probability score >0.73) | 65/117 (55.6) | 45/499 (9.0) | 59.1 (49.7-67.9) | NA | 55.6 (46.5-63.9) | NA | 6.16 (4.55-8.59) |
| Procalcitonin measurement | |||||||
| BVU (<0.10 ng/mL) | 53/117 (45.3) | 332/499 (66.5) | NA | 86.2 (82.7-89.5) | NA | 66.5 (61.9-70.8) | 0.68 (0.54-0.82) |
| BU (0.10-0.24 ng/mL) | 27/117 (23.1) | 104/499 (20.8) | NA | 79.4 (71.7-85.6) | NA | 20.8 (17.7-24.7) | 1.11 (0.74-1.59) |
| BL (0.25-0.50 ng/mL) | 11/117 (9.4) | 30/499 (6.0) | 26.8 (15.0-42.8) | NA | 9.4 (5.1-15.9) | NA | 1.56 (0.75-3.02) |
| BVL (>0.50 ng/mL) | 26/117 (22.2) | 33/499 (6.6) | 44.1 (30.4-57.1) | NA | 22.2 (15.4-30.5) | NA | 3.36 (2.06-5.24) |
Abbreviations: BL, bacterial likely; BU, bacterial unlikely; BVL, bacterial very likely; BVU, bacterial very unlikely; HR-B/V, host response bacterial/viral; NA, not applicable; NPV, negative predictive value; PPV, positive predictive value; VL, viral likely; VVL, viral very likely.
PPV and NPV were used for the diagnosis of bacterial infection.
Sensitivity and specificity were calculated as the percentage of adjudicated participants in each band. The test’s overall sensitivity and specificity were calculated as the sum of values in the respective cells. For example, the HR-B/V test had a sensitivity of 89.8% (53.1% plus 36.7%) for results in the BVL or BL groups and a specificity of 82.2% (60.4% plus 21.8%) for results in the VVL or VL groups.
Likelihood ratios were used for the diagnosis of bacterial infection. Values in the BVL and BL groups correspond to positive likelihood ratios, whereas values in the VL and VVL correspond to negative likelihood ratios.
Procalcitonin algorithms were also applied to stratify results into 4 likelihood categories (Figure 2A). The cohort was heavily skewed to the groups with low levels of procalcitonin: only 14 participants (28.6%) with bacterial infection were categorized in either of the 2 bacterial diagnostic groups (4 individuals in the bacteria likely group and 10 individuals in the bacteria very likely group) (Table 3; Figure 2C). For participants with procalcitonin levels lower than 0.10 ng/mL (bacterial very unlikely group), the NPV for bacterial infection was 89.5% (95% CI, 84.6%-93.1%). For participants with procalcitonin levels higher than 0.50 ng/mL (bacterial very likely group), we observed a PPV of 29.4% (95% CI, 15.9%-45.6%) for bacterial infection. These values were lower than those observed for the HR-B/V test (P < .001 for both comparisons).
Secondary Analysis
We next measured test performance among all 616 participants (499 with viral infection and 117 with bacterial infection), including those with a low-confidence adjudication (those without confirmatory microbiological test results or with discordant adjudications) for whom the reference standard was expected to be less accurate. In the full cohort, the HR-B/V test had sensitivity of 86.4% (95% CI, 79.6%-92.5%) and specificity of 71.9% (95% CI, 67.7%-75.9%) when applying a single threshold to discriminate bacterial from viral infection (Figure 2D; Table 2). As observed in the primary analysis, procalcitonin measurement was less sensitive (31.6%; 95% CI, 23.3%-39.5%; P < .001) but more specific (87.3%; 95% CI, 84.5%-90.1%; P = .005) than the HR-B/V test (Figure 2E), with an NPV of 84.5% (95% CI, 82.8%-86.1%) for bacterial infection. The procalcitonin area under the receiver operating characteristic curve was 0.62 (95% CI, 0.56-0.67) compared with 0.85 (95% CI, 0.81-0.89) for the HR-B/V test (P < .001) (eFigure in Supplement 1). When considering all 4 HR-B/V likelihood groups, the viral very likely group had an NPV of 96.7% (95% CI, 94.2%-98.4%), whereas the bacterial very likely group had a PPV of 59.1% (95% CI, 49.7%-67.9%) for bacterial infection. Additional test characteristics are shown in Table 3. A procalcitonin level of lower than 0.10 ng/mL had an NPV of 86.2% (95% CI, 82.7%-89.5%) for bacterial infection (P < .001 vs the HR-B/V test). A procalcitonin level higher than 0.50 ng/mL had a PPV of 44.1% (95% CI, 30.4%-57.1%) for bacterial infection (P = .005 vs the HR-B/V test).
Test Performance in Clinical Subgroups
We did not observe any significant differences in HR-B/V test performance based on age, enrollment site, race, ethnicity, or sex in either the primary or secondary analysis (eTable 2 in Supplement 1). Performance was unchanged by the presence of comorbidities, including coronary artery disease, heart failure, chronic kidney disease, chronic lung disease, diabetes, or the composite Charlson Comorbidity Index score. Using hospitalization as a surrogate for illness severity, we observed a higher overall accuracy among hospitalized vs nonhospitalized participants (78.2% vs 73.2%; P = .002). Treatment vs nontreatment with antibacterial therapy 8 hours or more before sample collection did not impact test performance as measured by overall accuracy (75.9% vs 74.5%; P = .73). Duration of illness, assessed in daily increments, also did not impact test performance (eg, 0-1 days vs 7 days since symptom onset: 76.6% vs 77.3%; P = .55). The test was less accurate among participants with enterovirus or rhinovirus (59.3%; 95% CI, 49.9%-68.3%) compared with participants with all other viral infections (90.2%; 95% CI, 85.9%-93.6%; P < .001) (eTable 3 in Supplement 1).
Antibacterial Therapy Use
This study did not directly measure the impact of HR-B/V testing on antibacterial therapy use because testing was not performed in real time. We estimated the potential association if results had been available. In the primary analysis involving 334 participants, 234 individuals were adjudicated as having viral infections, 84 of whom (35.9%) were prescribed antibacterial therapy. These 84 participants represented opportunities for reduced antibacterial medication use. In contrast, the HR-B/V test misclassified 51 participants as having bacterial infections, 30 of whom (58.8%) were not prescribed antibacterial therapy. Therefore, adherence to HR-B/V test results would have had the net result of eliminating inappropriate antibacterial therapy use in 54 of 334 cases (16.2%).
Classification of COVID-19 Cases
Among 33 participants with acute COVID-19, the HR-B/V test correctly classified 30 participants (90.9%) as having a viral infection (24 in the viral very likely group and 6 in the viral likely group). There was no apparent association between HR-B/V test results and participant-reported symptom severity (Figure 2F). In particular, the symptom severities were similar among the 3 misclassified participants compared with those who were correctly classified as having a viral infection.
Discussion
This diagnostic study found an NPV of up to 97.9% for bacterial infection using a rapid host gene expression test to discriminate bacterial from viral infection among a large multicenter heterogeneous pediatric and adult population. Antimicrobial resistance is currently increasing at a concerning rate, owing in part to inappropriate use of antibacterial therapy.6 Moreover, antibacterial medications pose risks to the individual patient, including allergic reactions, drug-drug interactions, and Clostridium difficile infections.25,26 To address this diagnostic challenge, we developed a test that discriminated bacterial from viral infection by measuring the host’s gene expression response. This study validated the performance of a research-use-only HR-B/V test for the discrimination of bacterial vs viral infection among 616 participants with febrile ARI. The gene expression test was superior to measurement of procalcitonin, a widely used host response peptide biomarker that has been validated among patients with lower respiratory tract infections. The test also worked equally well among participants with acute COVID-19. Although we did not assess the combined performance of the HR-B/V test and procalcitonin measurement, a previous study27 found no improvement vs HR-B/V testing alone.
The potential utility of a test distinguishing bacterial from viral infection is highly dependent on the prevalence of these conditions.18 When the prevalence of viral infection is high, as in most cases of ARI, a moderately accurate test would have a high NPV for bacterial infection. As a consequence, such a test does not need to have perfect accuracy to be clinically valuable. For example, despite 55% sensitivity and 76% specificity for procalcitonin measurement to detect bacterial pneumonia, the biomarker’s use has been reported to decrease unnecessary prescription of antibacterial therapy.13,15 However, this antibacterial-sparing outcome was not observed in the Procalcitonin Antibiotic Consensus Trial,14 a randomized clinical trial of procalcitonin measurement compared with standard of care among patients in the emergency department with lower respiratory tract infections. The reasons for this finding are likely multifactorial, although an insufficient NPV is a likely factor. Compared with procalcitonin measurement, the significantly higher NPV for the HR-B/V test (97.9%) suggests it may have even greater clinical utility, which will need to be verified in a prospective clinical utility study. The most helpful use of the test would therefore be a result indicating the absence of bacterial infection, especially in settings in which the prevalence of such infections is low, as it is with ARI cases. In the present study, we observed a potential reduction of at least 16.2%, even after accounting for false bacterial infection diagnoses in which antibacterial therapy might have been overprescribed. This reduction did not account for scenarios in which clinicians may have chosen to nevertheless prescribe antibacterial therapy, resulting in a lower effect size than we found possible. The test’s potential value will also depend on how easily it can be integrated into clinical workflows.
High rates of antibacterial medication overuse have occurred during the COVID-19 pandemic because of uncertainty about the presence of bacterial infection.6,7 Although the HR-B/V test was developed before the pandemic, participants with COVID-19 were identified as having viral infection without a concurrent bacterial infection, offering a tool to support antimicrobial stewardship. We were unable to identify cases of SARS-CoV-2 and bacterial coinfection; therefore, we could not draw any conclusions about test performance in this scenario, although this is an important clinical group to investigate. This caveat applies to cases of bacterial and viral coinfection in general, which were inadequately represented in the present cohort. However, a previous study16 found that the HR-B/V test signature identified bacterial infection in 100% of microbiologically confirmed bacterial and viral coinfection. It has also previously been reported that host gene expression can distinguish bacterial, viral, and noninfectious illness in patients with immunocompromising conditions.28
One challenge when evaluating tests for bacterial vs viral discrimination is the absence of a diagnostic criterion standard. Adjudication has often been used, with 1 study29 reporting reasonable interobserver agreement (κ = 0.88). However, the reliability of this reference decreases in several situations. For example, the absence of microbiological confirmation or discordant adjudications has been associated with lower test accuracy for other biomarkers.30 The present study revealed similar findings in which test accuracy decreased among participants with a low-confidence adjudication, defined as having no supportive microbiological results or discordant adjudications (secondary analysis cohort). This decrease was not likely to be associated with inadequate test performance but rather with errors in the reference standard used.
The development and validation of this research-use-only HR-B/V test measured using the BioFire Diagnostics system has been previously described.16 The current study confirms those findings while encompassing a broader age range, multiple sites, and a more racially and ethnically heterogeneous cohort. We did not observe any significant differences owing to sex, age, race, ethnicity, enrollment site, comorbidity, or hospitalization status. A previous study18 evaluating this test included patients with noninfectious illness. Although inclusion of these patients increased the population available for testing, it came at the expense of test accuracy. In particular, test performance decreased as the number of clinical categories increased.18,31 Therefore, the present study focused on the narrower indication of acute (≤7 days) febrile respiratory illness suspected to be caused by bacterial or viral infection.
Measuring gene expression simply, rapidly, and reliably requires the convergence of multiple technologies, including blood cell lysis, RNA purification, RNA preservation or rapid processing, and quantitative real-time reverse transcription amplification (emerging technologies could substitute for some of these elements). This process has been conducted using slower high-complexity research tools, such as low-density array cards (TaqMan; Thermo Fisher Scientific) or nanostring arrays (NanoString Technologies).28,32,33,34 There has been more limited progress in the development of rapid sample-to-answer tests. Sutherland et al17 described a host gene expression test for the diagnosis of tuberculosis using a molecular analyzer (GeneXpert System; Cepheid). Other studies have described the successful translation of host gene expression tests for viral infection and bacterial vs viral discrimination using 2 different platforms (Qvella Corporation and BioFire Diagnostics, respectively).16,35 This previous work included participants with bacterial and viral coinfection, finding the ability to detect host responses to both pathogens. The present study extended those findings, highlighting the opportunity to develop host gene expression–based diagnostic assessments for a variety of clinical applications. Signatures have been described for infectious disease applications (eg, sepsis, candidemia, Dengue, tuberculosis, and malaria)36,37,38,39,40,41,42,43,44,45,46,47 and noncommunicable diseases (eg, rheumatic diseases, coronary artery disease, radiation exposure, and cancer).48,49,50,51,52,53 As the path from signature discovery to test development and commercialization comes into focus, we expect the development and use of point-of-need host response–based diagnostic tools to expand.
Limitations
This study has several limitations. These limitations include the use of clinical adjudication as an imperfect reference standard. This study did not address elements of analytical validation, such as site-to-site or run-to-run variability, nor did it evaluate patients with nonrespiratory infections. Furthermore, tests that discriminate bacterial from viral infection are only useful if clinicians modify their antibacterial medication use accordingly. Although we found a reduction in the use of antibacterial therapy is possible, this study did not address clinical utility. We did not identify demographic or clinical variables associated with test performance. Although hospitalization as a marker of severity did not impact test performance, there were too few participants with critical illness defined by intensive care unit admission to evaluate the test in this subgroup. There were also too few participants with bacterial and viral coinfection to compare results with those of other published reports. Serial HR-B/V test measurements would help to define a role for the test in treatment-response monitoring.
Conclusions
This diagnostic study found that host gene expression could be measured at the point of need using a readily available clinical platform. The gene expression biomarkers included in the HR-B/V test accurately distinguished bacterial from viral infections among participants with acute febrile respiratory tract infections. Test performance was superior to procalcitonin measurement and was also accurate among participants with COVID-19.
eMethods. Study Design and Reference Standard
eTable 1. Distribution of Participants in Primary and Secondary Analyses by Adjudication Confidence and Microbiological Cause
eTable 2. HR-B/V Test Performance Stratified by Demographic or Clinical Parameters
eTable 3. Microbiological Causes and Associated HR-B/V Test Accuracies
eFigure. Area Under the Receiver Operating Characteristic Curves for HR-B/V Test and Procalcitonin Measurement in the Primary Analysis and Secondary Analysis Cohorts
Nonauthor Collaborators
References
- 1.Centers for Disease Control and Prevention . National Ambulatory Medical Care Survey: 2015 state and national summary tables. Centers for Disease Control and Prevention; 2015. Accessed December 5, 2018. https://www.cdc.gov/nchs/data/ahcd/namcs_summary/2015_namcs_web_tables.pdf
- 2.GBD 2015 Disease and Injury Incidence and Prevalence Collaborators . Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet. 2016;388(10053):1545-1602. doi: 10.1016/S0140-6736(16)31678-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Shapiro DJ, Hicks LA, Pavia AT, Hersh AL. Antibiotic prescribing for adults in ambulatory care in the USA, 2007-09. J Antimicrob Chemother. 2014;69(1):234-240. doi: 10.1093/jac/dkt301 [DOI] [PubMed] [Google Scholar]
- 4.Lee GC, Reveles KR, Attridge RT, et al. Outpatient antibiotic prescribing in the United States: 2000 to 2010. BMC Med. 2014;12:96. doi: 10.1186/1741-7015-12-96 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Havers FP, Hicks LA, Chung JR, et al. Outpatient antibiotic prescribing for acute respiratory infections during influenza seasons. JAMA Netw Open. 2018;1(2):e180243. doi: 10.1001/jamanetworkopen.2018.0243 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Ansari S, Hays JP, Kemp A, et al. ; Global AMR Insights Ambassador Network . The potential impact of the COVID-19 pandemic on global antimicrobial and biocide resistance: an AMR Insights global perspective. JAC Antimicrob Resist. 2021;3(2):dlab038. doi: 10.1093/jacamr/dlab038 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Al-Hadidi SH, Alhussain H, Hadi HA, et al. The spectrum of antibiotic prescribing during COVID-19 pandemic: a systematic literature review. Microb Drug Resist. 2021;27(12):1705-1725. doi: 10.1089/mdr.2020.0619 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Jain S, Self WH, Wunderink RG, et al. ; CDC EPIC Study Team . Community-acquired pneumonia requiring hospitalization among U.S. adults. N Engl J Med. 2015;373(5):415-427. doi: 10.1056/NEJMoa1500245 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Shaman J, Morita H, Birger R, et al. Asymptomatic summertime shedding of respiratory viruses. J Infect Dis. 2018;217(7):1074-1077. doi: 10.1093/infdis/jix685 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Martin JM, Green M, Barbadora KA, Wald ER. Group A streptococci among school-aged children: clinical characteristics and the carrier state. Pediatrics. 2004;114(5):1212-1219. doi: 10.1542/peds.2004-0133 [DOI] [PubMed] [Google Scholar]
- 11.Self WH, Balk RA, Grijalva CG, et al. Procalcitonin as a marker of etiology in adults hospitalized with community-acquired pneumonia. Clin Infect Dis. 2017;65(2):183-190. doi: 10.1093/cid/cix317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Musher DM, Bebko SP, Roig IL. Serum procalcitonin level, viral polymerase chain reaction analysis, and lower respiratory tract infection. J Infect Dis. 2014;209(4):631-633. doi: 10.1093/infdis/jit579 [DOI] [PubMed] [Google Scholar]
- 13.Kamat IS, Ramachandran V, Eswaran H, Guffey D, Musher DM. Procalcitonin to distinguish viral from bacterial pneumonia: a systematic review and meta-analysis. Clin Infect Dis. 2020;70(3):538-542. doi: 10.1093/cid/ciz545 [DOI] [PubMed] [Google Scholar]
- 14.Huang DT, Yealy DM, Filbin MR, et al. ; ProACT Investigators . Procalcitonin-guided use of antibiotics for lower respiratory tract infection. N Engl J Med. 2018;379(3):236-249. doi: 10.1056/NEJMoa1802670 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Schuetz P, Wirz Y, Sager R, et al. Procalcitonin to initiate or discontinue antibiotics in acute respiratory tract infections. Cochrane Database Syst Rev. 2017;10:CD007498. doi: 10.1002/14651858.CD007498.pub3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tsalik EL, Henao R, Montgomery JL, et al. ; Antibacterial Resistance Leadership Group . Discriminating bacterial and viral infection using a rapid host gene expression test. Crit Care Med. 2021;49(10):1651-1663. doi: 10.1097/CCM.0000000000005085 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Sutherland JS, van der Spuy G, Gindeh A, et al. ; TrENDx-TB Consortium . Diagnostic accuracy of the Cepheid 3–gene host response fingerstick blood test in a prospective, multi-site study: interim results. Clin Infect Dis. Published online September 22, 2021. doi: 10.1093/cid/ciab839 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Tsalik EL, Henao R, Nichols M, et al. Host gene expression classifiers diagnose acute respiratory illness etiology. Sci Transl Med. 2016;8(322):322ra11. doi: 10.1126/scitranslmed.aad6873 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bossuyt PM, Reitsma JB, Bruns DE, et al. ; STARD Group . STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ. 2015;351:h5527. doi: 10.1136/bmj.h5527 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-383. doi: 10.1016/0021-9681(87)90171-8 [DOI] [PubMed] [Google Scholar]
- 21.Fawcett T. An introduction to ROC analysis. Pattern Recognit Lett. 2006;27(8):861-874. doi: 10.1016/j.patrec.2005.10.010 [DOI] [Google Scholar]
- 22.Christ-Crain M, Jaccard-Stolz D, Bingisser R, et al. Effect of procalcitonin-guided treatment on antibiotic use and outcome in lower respiratory tract infections: cluster-randomised, single-blinded intervention trial. Lancet. 2004;363(9409):600-607. doi: 10.1016/S0140-6736(04)15591-8 [DOI] [PubMed] [Google Scholar]
- 23.Efron B. Better bootstrap confidence intervals. J Am Stat Assoc. 1987;82(397):171-185. doi: 10.1080/01621459.1987.10478410 [DOI] [Google Scholar]
- 24.DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44(3):837-845. doi: 10.2307/2531595 [DOI] [PubMed] [Google Scholar]
- 25.Shehab N, Lovegrove MC, Geller AI, Rose KO, Weidle NJ, Budnitz DS. US emergency department visits for outpatient adverse drug events, 2013-2014. JAMA. 2016;316(20):2115-2125. doi: 10.1001/jama.2016.16201 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Shehab N, Patel PR, Srinivasan A, Budnitz DS. Emergency department visits for antibiotic-associated adverse events. Clin Infect Dis. 2008;47(6):735-743. doi: 10.1086/591126 [DOI] [PubMed] [Google Scholar]
- 27.Ross M, Henao R, Burke TW, et al. A comparison of host response strategies to distinguish bacterial and viral infection. PLoS One. 2021;16(12):e0261385. doi: 10.1371/journal.pone.0261385 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Mahle RE, Suchindran S, Henao R, et al. Validation of a host gene expression test for bacterial/viral discrimination in immunocompromised hosts. Clin Infect Dis. 2021;73(4):605-613. doi: 10.1093/cid/ciab043 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.van Houten CB, Naaktgeboren CA, Ashkenazi-Hoffnung L, et al. ; IMPRIND Consortium . Expert panel diagnosis demonstrated high reproducibility as reference standard in infectious diseases. J Clin Epidemiol. 2019;112:20-27. doi: 10.1016/j.jclinepi.2019.03.010 [DOI] [PubMed] [Google Scholar]
- 30.Srugo I, Klein A, Stein M, et al. Validation of a novel assay to distinguish bacterial and viral infections. Pediatrics. 2017;140(4):e20163453. doi: 10.1542/peds.2016-3453 [DOI] [PubMed] [Google Scholar]
- 31.Bodkin N, Ross M, McClain MT, et al. Systematic comparison of published host gene expression signatures for bacterial/viral discrimination. Genome Med. 2022;14(1):18. doi: 10.1186/s13073-022-01025-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Mayhew MB, Buturovic L, Luethy R, et al. A generalizable 29-mRNA neural-network classifier for acute bacterial and viral infections. Nat Commun. 2020;11(1):1177. doi: 10.1038/s41467-020-14975-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Lydon EC, Bullard C, Aydin M, et al. A host gene expression approach for identifying triggers of asthma exacerbations. PLoS One. 2019;14(4):e0214871. doi: 10.1371/journal.pone.0214871 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Lydon EC, Henao R, Burke TW, et al. Validation of a host response test to distinguish bacterial and viral respiratory infection. EBioMedicine. 2019;48:453-461. doi: 10.1016/j.ebiom.2019.09.040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Tsalik EL, Khine A, Talebpour A, et al. Rapid, sample-to-answer host gene expression test to diagnose viral infection. Open Forum Infect Dis. 2019;6(11):ofz466. doi: 10.1093/ofid/ofz466 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Tsalik EL, Langley RJ, Dinwiddie DL, et al. An integrated transcriptome and expressed variant analysis of sepsis survival and death. Genome Med. 2014;6(11):111. doi: 10.1186/s13073-014-0111-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Sweeney TE, Perumal TM, Henao R, et al. A community approach to mortality prediction in sepsis via gene expression analysis. Nat Commun. 2018;9(1):694. doi: 10.1038/s41467-018-03078-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.DeMerle KM, Angus DC, Baillie JK, et al. Sepsis Subclasses: A Framework for Development and Interpretation. Crit Care Med. 2021;49(5):748-759. doi: 10.1097/CCM.0000000000004842 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Miller RR III, Lopansri BK, Burke JP, et al. Validation of a host response assay, SeptiCyte LAB, for discriminating sepsis from systemic inflammatory response syndrome in the ICU. Am J Respir Crit Care Med. 2018;198(7):903-913. doi: 10.1164/rccm.201712-2472OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Zimmerman JJ, Sullivan E, Yager TD, et al. Diagnostic accuracy of a host gene expression signature that discriminates clinical severe sepsis syndrome and infection-negative systemic inflammation among critically ill children. Crit Care Med. 2017;45(4):e418-e425. doi: 10.1097/CCM.0000000000002100 [DOI] [PubMed] [Google Scholar]
- 41.Steinbrink JM, Myers RA, Hua K, et al. The host transcriptional response to Candidemia is dominated by neutrophil activation and heme biosynthesis and supports novel diagnostic approaches. Genome Med. 2021;13(1):108. doi: 10.1186/s13073-021-00924-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Nikolayeva I, Bost P, Casademont I, et al. A blood RNA signature detecting severe disease in young Dengue patients at hospital arrival. J Infect Dis. 2018;217(11):1690-1698. doi: 10.1093/infdis/jiy086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Robinson M, Sweeney TE, Barouch-Bentov R, et al. A 20-gene set predictive of progression to severe Dengue. Cell Rep. 2019;26(5):1104-1111. doi: 10.1016/j.celrep.2019.01.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Valim C, Ahmad R, Lanaspa M, et al. Responses to bacteria, virus, and malaria distinguish the etiology of pediatric clinical pneumonia. Am J Respir Crit Care Med. 2016;193(4):448-459. doi: 10.1164/rccm.201506-1100OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Bloom CI, Graham CM, Berry MPR, et al. Transcriptional blood signatures distinguish pulmonary tuberculosis, pulmonary sarcoidosis, pneumonias and lung cancers. PLoS One. 2013;8(8):e70630. doi: 10.1371/journal.pone.0070630 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Mihret A, Loxton AG, Bekele Y, et al. Combination of gene expression patterns in whole blood discriminate between tuberculosis infection states. BMC Infect Dis. 2014;14:257. doi: 10.1186/1471-2334-14-257 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Singhania A, Verma R, Graham CM, et al. A modular transcriptional signature identifies phenotypic heterogeneity of human tuberculosis infection. Nat Commun. 2018;9(1):2308. doi: 10.1038/s41467-018-04579-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Voora D, Coles A, Lee KL, et al. An age- and sex-specific gene expression score is associated with revascularization and coronary artery disease: insights from the Prospective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial. Am Heart J. 2017;184:133-140. doi: 10.1016/j.ahj.2016.11.004 [DOI] [PubMed] [Google Scholar]
- 49.Voora D, Cyr D, Lucas J, et al. Aspirin exposure reveals novel genes associated with platelet function and cardiovascular events. J Am Coll Cardiol. 2013;62(14):1267-1276. doi: 10.1016/j.jacc.2013.05.073 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Dressman HK, Muramoto GG, Chao NJ, et al. Gene expression signatures that predict radiation exposure in mice and humans. PLoS Med. 2007;4(4):e106. doi: 10.1371/journal.pmed.0040106 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Glas AM, Floore A, Delahaye LJMJ, et al. Converting a breast cancer microarray signature into a high-throughput diagnostic test. BMC Genomics. 2006;7:278. doi: 10.1186/1471-2164-7-278 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Baechler EC, Batliwalla FM, Karypis G, et al. Interferon-inducible gene expression signature in peripheral blood cells of patients with severe lupus. Proc Natl Acad Sci U S A. 2003;100(5):2610-2615. doi: 10.1073/pnas.0337679100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Chaussabel D, Allman W, Mejias A, et al. Analysis of significance patterns identifies ubiquitous and disease-specific gene-expression signatures in patient peripheral blood leukocytes. Ann N Y Acad Sci. 2005;1062:146-154. doi: 10.1196/annals.1358.017 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Study Design and Reference Standard
eTable 1. Distribution of Participants in Primary and Secondary Analyses by Adjudication Confidence and Microbiological Cause
eTable 2. HR-B/V Test Performance Stratified by Demographic or Clinical Parameters
eTable 3. Microbiological Causes and Associated HR-B/V Test Accuracies
eFigure. Area Under the Receiver Operating Characteristic Curves for HR-B/V Test and Procalcitonin Measurement in the Primary Analysis and Secondary Analysis Cohorts
Nonauthor Collaborators