Impact of Rapid Molecular Respiratory Virus Testing on Real-Time Decision Making in a Pediatric Emergency Department

Abstract

Acute respiratory illnesses (ARIs) are usually viral [influenza, respiratory syncytial virus (RSV)] and account for 25% of emergency department (ED) peak-season visits. Laboratory PCR testing is accurate albeit slow, whereas rapid antigen testing is inaccurate. We determined the impact of bedside PCR (molecular point-of-care test; mPOCT) on pediatric ARI management. This was a prospective cohort study of consecutive pediatric patients with ED-ordered respiratory PCR test, enrolled over 9 weeks during peak flu season. On ordering, ED physicians were interviewed to ascertain real-time plans if given immediate influenza/RSV PCR results for the current patient. Two groups were compared: actual management and management adjusted for mPOCT results. We compared ED length of stay (LOS), tests ordered, and antibiotic/antiviral ordering. One-hundred thirty-six respiratory PCR panels were ordered, 71 by admitting team, 61 for ED management. Of 61 ED-initiated tests, physicians indicated in 39 cases (64%) they would change patient management were bedside viral results available. Physicians would have decreased ED LOS by 33 minutes, ordered fewer tests (18%; P < 0.001) with average patient charge savings of $669, fewer antibiotics among discharged patients (17%; P = 0.043), and increased appropriate antiviral use (13%; P = 0.023). Rapid bedside ARI mPOCT PCR has the potential to decrease ED LOS, reduce diagnostic tests and patient charges, and increase appropriate use of antibiotics and antiviral agents.

During the winter months, fever and respiratory infection symptoms make up to 25% of all emergency department (ED) visits.¹ Acute respiratory illness (ARI) is a leading cause of hospitalization for young children, contributing to 10.4% of all deaths in children younger than 5 years.2, 3 ARI has a large spectrum of disease, ranging from mild upper respiratory tract problems to severe lower respiratory infections (eg, bronchiolitis and pneumonia) that can be associated with significant rates of morbidity and mortality. Although most ARIs are viral in cause with influenza (A and B) and respiratory syncytial virus (RSV) being most common, symptoms are often nonspecific, therefore making causative diagnosis based on clinical presentation unreliable.⁴ Furthermore, the Centers for Disease Control and Prevention currently recommend administration of antiviral agents within 48 hours of symptom onset for children younger than 2 years of age or young immunocompromised children who are at high risk of influenza-related complications.⁵

There is a need and desire to improve diagnosis and management in the ED setting.⁶ Current criterion standard laboratory tests based on traditional real-time PCR, which may require up to several hours for turnaround time, are too slow to affect ED management.⁷ Without a confirmed viral diagnosis, ED physicians may resort to precautionary patient management strategies that result in antibiotic overuse and antiviral misuse, additional diagnostic testing, and unnecessary hospitalizations requiring isolation beds.8, 9 These conservative measures promote antibiotic and antiviral resistance in the population and increase overall health system expenditures. In particular they contribute to prolonged ED wait times, length of stay (LOS), and overcrowding. Further, prior studies have shown that reducing testing turnaround times and initiating diagnostic testing earlier during ED triage reduces ED LOS.10, 11 Antigen tests can provide results between 30 and 150 minutes with near-patient testing capability, but they have unacceptable sensitivity as low as 10% for influenza and RSV in certain studies.12, 13 As a result, studies have shown antigen testing has limited impact on ED patient management.¹⁴

Recent technologic advances in molecular diagnostics have enabled the development of fully automated PCR platforms with point-of-care (POC) capability to detect influenza A and B and RSV with >95% sensitivity and specificity and turnaround time as fast as 20 minutes.15, 16, 17, 18 These emerging rapid molecular POC tests (mPOCTs) are designed to be performed at the bedside by minimally trained personnel. Before the clinical availability of these tests with Food and Drug Administration (FDA) clearance for waived status under Clinical Laboratory Improvement Amendments (CLIA), we performed a study to determine the impact and potential value of rapid influenza and RSV PCR results on physician decision making in a pediatric ED during peak ARI season (https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K153544; http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfCLIA/Detail.cfm?ID=39763&NoClia=1, last accessed January 10, 2017).

Materials and Methods

Study Design

This was a prospective observational study with real-time interviews of physicians during active patient management in the ED, when a PCR test for respiratory viruses was ordered.

Study Setting

The study occurred in the pediatric ED of an academic medical center during peak ARI season. It involved consecutive pediatric patients younger than 18 years of age who had a respiratory virus PCR panel by nasopharyngeal swab in the Pediatric Emergency Department at Stanford University Medical Center during the 9-week study period from January 10, 2016, to March 13, 2016.

Study Protocol

During the study period patients were identified by a real-time electronic notification system developed to identify patients in real time for clinical studies.¹⁹ No post hoc convenience surveys were administered at any time. The electronic notification was set according to the order coming from the pediatric ED in a patient younger than 18 months for a respiratory virus PCR panel. This set up a real-time notification to the on-call research coordinator who then contacted the ordering attending ED physician to conduct a brief survey relating to patient management within minutes of the respiratory panel order being placed. Given the real-time nature of the electronic notifications and immediate subsequent interviews, interviews were possible at any time of day throughout the study period.

At the time of the survey, while patients were still being actively managed in the ED, physicians were informed that the viral PCR as an mPOCT would have results within 20 minutes of a nasopharyngeal swab. Further, physicians were informed the test would present individually positive or negative viral presence results for RSV, influenza A, and influenza B (influenza A and B collectively referred to as influenza). This theoretical mPOCT was considered to have the same diagnostic accuracy as the commercially available standard respiratory panel PCR test used at the institution (Respiratory Virus Panel XT8; GenMark, Fremont, CA), albeit only testing for RSV and influenza A/B.¹⁵ Physicians were asked hypothetically how their patient management would change if the mPOCT results were available imminently, including whether fewer diagnostic testing [urinalysis (UA), blood draw, or chest X-ray (CXR)] would have been pursued if a source of fever was identified (Supplemental Figure S1). Potential changes in antibiotic use and oseltamivir use and changes in disposition were also surveyed. Physicians' a priori proposed plans according to potential mPOCT results were retroactively aligned with test results from the hospital laboratory standard 14-virus PCR test; this allowed determination of individual theoretical management plans that the physician would have followed if test results had been known in the ED compared with actual ED management performed in the absence of test result information.

Hospital Laboratory Respiratory Viral Detection

Per standard practice at our institution, ED nasopharyngeal swabs for viral testing are transported to an off-site institutional facility for processing by fully trained laboratory staff. Viral DNA/RNA is extracted with the EZ1 Virus Mini Kit version 2.0 (Qiagen, Hilden, Germany), and virus is detected via Respiratory Virus Panel XT8 (GenMark). Total turnaround time is between 8 and 24 hours, factoring in transport and handling, assay time (7 hours), and allowance for batch testing (two to five times daily, seasonally depending on staffing and volume concerns). No changes in institutional standard of practice for ordering or processing respiratory virus panels were made during the conduction of this study.

Measurements of Key Outcomes

The primary outcome was the potential change in ED LOS. Patient-specific outcomes, including ED LOS, were compared as baseline (true encounter) versus test-dependent management (hypothetical management) according to the potential ED LOS from the assumption that the POC test would have the actual results of the institution standard PCR test that was performed. For cases when physicians may have changed management by altering the number of ancillary diagnostic studies performed, we compared the ED LOS for two groups: actual ED LOS versus ED LOS if mPOCT was available (adjusted LOS). Baseline ED LOS was calculated from triage time to selection of disposition, reflecting the evaluation time by the ED physician rather than logistics of leaving the ED according to the disposition. For the adjusted ED LOS group, we calculated the ED LOS for these cases according to the actual results and assuming the mPOCT would have the same sensitivity and specificity as the gold standard test. In cases when mPOCT availability would have resulted in changes in ED management (reduced number of ancillary tests), the triage-to-disposition time was changed to the triage-to-respiratory panel order time plus 20 minutes to allow time for the test to result.

Secondary outcomes included admission status, patient charges for additional diagnostic evaluations for fever source (UA, blood draw, CXR), and both antibacterial and antiviral prescription incidence. Demographic and encounter-specific medical information was obtained by chart review and administrative databases, including actual charges for the assessed procedures that included standard UA screen (protocol for culture if positive), blood draw with culture and complete blood count, and two-view CXR.

Sample Size and Data Analysis

We determined that we would need to have 28 patients serving as their own comparison (ie, 28 in each group) to determine the effect on the primary outcome of ED LOS. This was calculated using an α of 0.05 and power of 80% to determine 10% decrease in ED LOS of 20 minutes (average 200 minutes) as an important decrease in ED LOS.

The data were considered paired, and as such, continuous outcomes were measured with paired t-tests and dichotomous outcomes with McNemar's test as appropriate. Statistical analyses were performed using GraphPad Prism version 6 (GraphPad Inc., La Jolla, CA) and reported as P values and 95% CIs when appropriate.

Institutional Review Board Considerations

There were no extra tests or procedures and no change in the care for patients, and the institutional review board waived consent for patients who met inclusion criteria. Attending physicians who would be completing the surveys received in-service instruction and were required to receive and acknowledge an e-mail message detailing the study and their ability to refuse participation and contact the institutional review board with concerns before onset of the study. No physicians refused to participate.

Results

During the study period, 136 respiratory PCR panels were ordered by 25 unique attending physicians staffing the pediatric ED (Figure 1); of these, 132 panels were completed, with 4 being incomplete because of patient declination of nasopharyngeal swab or patient leaving the ED before sample collection. Seventy-one were ordered on behalf of the admitting pediatric team, whereas 61 panels were ordered for ED management. Of these 61 ED physician-initiated tests, the treating physician at the time of patient management stated that in 39 cases (64%) they would consider changing their patient management strategy if the influenza or RSV results were immediately available. Patients for whom management would be altered with immediately available test results were predominantly younger than 5 years (Table 1). In particular, physicians would have altered their management by forgoing further testing (UA, blood draw, or CXR) if positive virus presence was confirmed by mPOCT, changing therapeutic decisions, or altering disposition. Table 2 summarizes the potential management changes had the results been immediately available. The effect of these changes would have decreased the mean ED LOS by 33 minutes (95% CI, 12–54 minutes) in this group, with mean baseline ED LOS of 212 minutes versus mean adjusted ED LOS of 179 minutes.

Figure 1.

Study participants. Of the four panels not completed, for one encounter the physician completed the full survey [ordered for emergency department (ED) management, would consider changing management]. For survey-based, outcomes such as determining baseline intent to change management, these data were included for total n = 40; for actual versus theoretical comparisons, these data were excluded, given there was no panel conducted to determine viral status. ICU, intensive care unit.

Table 1.

Demographic Characteristics of Patients for Whom Management Decisions Are Affected by Test Results (n = 39)

Characteristic	Value
Age, mean ± SD, years	3.8 ± 4.2
Age group, %
<3 months	10.3
3 to <6 months	12.8
6 to <12 months	12.8
1 to <5 years	30.8
5 to <10 years	23.1
10 to <18 years	10.3
Female sex, %	35.9
Race, %
White	12.8
Asian/Pacific Islander	30.8
Hispanic	48.7
Black	5.1
Other	2.6
ED presentation
Febrile (temperature ≥38.0°C), %	61.5
Tachycardia∗, %	92.3
Tachypnea∗, %	33.3
Admitted, %	41.0

Admission includes general floor, isolation, intensive care unit; no patients were admitted for observation (<24 hours) or to a clinical decision unit.

ED, emergency department.

^∗Age-related vitals.

Table 2.

Survey Responses: Physicians Indicate Their Management Plans according to Different Test Results

Management decision	RSV, % (95% CI)			Influenza, % (95% CI)			Summative, % (95% CI)
	Pos. (+) (n = 40)	Neg. (−) (n = 40)	P1	Pos. (+) (n = 40)	Neg. (−) (n = 40)	P2	Any (+) (n = 80)	Any (−) (n = 80)	P3	All results (n = 160)
ED diagnostics
Chest X-ray	28 (13–42)	53 (36–69)	0.001	35 (20–50)	53 (36–69)	0.007	31 (21–42)	53 (41–64)	<0.001	42 (34–50)
UA screen	23 (9–36)	35 (20–50)	0.023	15 (3–27)	40 (24–56)	0.001	19 (10–27)	38 (27–48)	<0.001	28 (21–35)
Blood draw	30 (15–45)	50 (34–66)	0.003	28 (13–42)	53 (36–69)	0.001	29 (19–39)	51 (40–62)	<0.001	40 (32–48)
Admission status
No change	80 (67–93)	88 (77–98)	0.083	90 (80–100)	90 (80–100)	>0.999	85 (77–93)	89 (82–96)	0.181	87 (82–92)
Discharge to admit	8 (−1 to 16)	5 (−2 to 12)	0.570	3 (−3 to 8)	5 (−2 to 12)	0.324	5 (0–10)	5 (0–10)	>0.999	5 (2–8)
Admit to discharge	5 (−2 to 12)	5 (−2 to 12)	>0.999	0 (0–0)	5 (−2 to 12)	0.160	3 (−1 to 6)	5 (0–10)	0.418	4 (1–7)
Total change	13 (2–23)	10 (0–20)	0.711	3 (−3 to 8)	10 (0–20)	0.083	8 (2–13)	10 (3–17)	0.531	9 (4–13)
Antimicrobial use
Antibiotics	18 (5–30)	15 (3–27)	0.744	18 (5–30)	25 (11–39)	0.262	18 (9–26)	20 (11–29)	0.620	19 (13–25)
Oseltamivir				85 (73–97)	10 (0–20)	<0.001

Values for P1 and P2 indicate comparison of management plans between positive and negative results of RSV and influenza, respectively; P3 compares management plans between a positive result from either RSV or influenza and a negative result from either test. Forty surveys were completed with indication that the treating physician would alter management according to test results; one patient did not have nasal swab completed for respiratory virus PCR testing, but survey results were retained, given it represents physician plans without regard to final panel results. Each survey allowed physicians to propose possible management plans for each of four scenarios (positive or negative, RSV, or influenza) for a total of n = 160 plans proposed from survey.

ED, emergency department; RSV, respiratory syncytial virus; UA, urinalysis.

Given scenarios for potential mPOCT results (either positive or negative for RSV or influenza), physicians proposed their diagnostic, therapeutic, and disposition plans (Table 2). Any positive viral result prompted a significant decrease in the number of diagnostic assays (CXR, UA, and blood tests) ordered by ED physicians; in general, tests were ordered a little more than half the time if viral results were negative, but they were ordered close to a quarter of the time if a positive result would have been returned. Admission status was largely unaffected by viral result. Oseltamivir prescriptions were significantly different between influenza-positive and influenza-negative patient management plans; most physicians would prescribe oseltamivir for patients positive for influenza (85%), and deferring oseltamivir for negative results (10%).

A variety of viruses were detected by the standard viral PCR panel during the study period (Table 3). The proposed in-ED mPOCT, with faster turnaround time, would enable the potential to identify most patients as having RSV, influenza, or no infection. With an 11-virus standard PCR panel, only 31% of patients had a viral infection from the nine viruses not covered by the proposed mPOCT. Twenty-nine percent of patients had no detectable virus, whereas a little >40% had either RSV or influenza infections, covering 69% of the tested population.

Table 3.

Respiratory Virus Detection by Institution Standard PCR Test during Study Period (n = 132)

Detection results	Prevalence
RSV (+), %	28.0
Influenza (+), %	13.6
Influenza A (+), %	2.3
Influenza B (+), %	11.4
Both RSV (+) and influenza (+), %	1.5
Virus (+), other than RSV or influenza	31.1
Number of viruses (+) on panel
0, %	28.8
1, %	56.1
2, %	13.6
3, %	0.8
4, %	0.8
≥5, %	0.0

All values were calculated as fraction of entire tested sample (n = 132). Subtypes of influenza A were not reported here.

+, positive test results; −, negative test results; RSV, respiratory syncytial virus.

Secondary outcomes for this study included patient charge savings, ancillary test use, disposition, antibiotic use, and oseltamivir use (Table 4). Substantial charge savings were seen in 12 of 39 patients with affected management plans, with $669 saved per patient with fewer tests (95% CI, $421–$918). Diagnostic testing fell by 12% to 18% when PCR test results were available earlier, although disposition was essentially unchanged. Physicians would have prescribed 17% fewer antibiotics for discharged patients if they would have had the results immediately available and would have been more likely to prescribe oseltamivir to patients testing positive and to withhold it in patients testing negative (13% net increase in appropriate oseltamivir prescriptions).

Table 4.

Changes in Secondary Outcomes if Survey Responses Reflected Changes in ED Management

Secondary outcome	Mean change (95% CI)	Patients affected, n
ED charge savings
Among all patients, $	−205.97 (−83.72 to −328.23)	39
Patients with changes only, $	−669.42 (−420.95 to −917.88)	12
ED diagnostics use changes
Chest X-ray, net %	−12.8 (−23.8 to −1.8)	7/39
UA screen, net %	−18.0 (−30.6 to −5.3)	5/39
Blood draw, net %	−12.8 (−23.8 to −1.8)	5/39
Any change, net %	−30.8 (−45.9 to −15.6)	12/39
Admission status changes
No change, %	94.9 (87.9–102.1)	37/39
Discharge to admit, %	0.0 (0.0–0.0)	0/39
Admit to discharge, %	5.1 (−2.1 to 12.4)	2/39
Total changes, %	5.1 (−2.1 to 12.4)	2/39
Antibiotic use changes
All patients, net %	−15.4 (−31.2 to 0.5)	10/39
Added antibiotics, %	5.1 (−2.1 to 12.4)	2/39
Reduced antibiotics, %	20.5 (7.3–33.8)	8/39
Discharged patients only, net %	−17.4 (−34.2 to −0.6)	4/23
Added antibiotics, %	0.0 (0.0–0.0)	0/23
Reduced antibiotics, %	17.4 (0.6−34.2)	4/23
Oseltamivir use changes
All patients, net %	7.7 (−6.0 to 21.4)	7/39
Added oseltamivir, %	12.8 (1.8–23.8)	5/39
Reduced oseltamivir, %	5.1 (−2.1 to 12.4)	2/39
Discharged patients only, net %	2.6 (−9.2 to 14.3)	5/23
Added oseltamivir, %	13.0 (−1.8 to 28.0)	3/23
Reduced oseltamivir, %	8.7 (−3.8 to 21.2)	2/23
Appropriate oseltamivir use
Pretest, total %	76.9 (63.1–90.8)	30/39
Prescribed (influenza positive), %	30.0 (−4.6 to 64.6)	3/10
Not prescribed (influenza negative), %	93.1 (83.3–102.9)	27/29
Posttest, total %	84.6 (72.8–96.5)	33/39
Prescribed (influenza positive), %	60.0 (23.1–96.9)	6/10
Not prescribed (influenza negative), %	93.1 (93.3–102.9)	27/29
Net change, %	7.7 (1.8–23.8)	3/39

Charge savings were calculated from per-test patient charged for chest X-ray, UA screen, and blood draw; patients with changes are those with changes in the number of diagnostic studies performed. For diagnostic studies, all (n = 12) affected patients had reductions in studies ordered; none of the patients would have had a management change that increased the number of diagnostic studies. Admission includes general floor, isolation, and intensive care; no enrolled patients were admitted for observation (<24 hours) or to a clinical decision unit. Discharged status from pretest disposition (n = 23), not posttest theoretical disposition (n = 22 discharged). Appropriate oseltamivir usage defined as oseltamivir prescription for influenza-positive patients or no oseltamivir prescription for influenza-negative patients.

ED, emergency department; UA, urinalysis.

Discussion

In this study we found that a rapid turnaround of influenza and RSV PCR test results, by mPOCT, would have affected most cases when a pediatric ED physician ordered nasopharyngeal swabs for respiratory viral PCR panels during the management of patients. In these cases mPOCT has the potential to improve ED LOS, decrease ordering of other tests, and improve the use of antibiotic and antiviral agents in these patients. The mPOCT would have had little if any impact on disposition of admitted patients.

Commercially available rapid respiratory viral PCR tests are becoming more prevalent. They far exceed previous antigen tests in terms of accuracy and speed.²⁰ Most are still awaiting FDA clearance or CLIA waiver for use outside of central laboratories. Two small studies assessed the clinical use of Verigene Respiratory Virus Plus test (Nanosphere, Inc., Northbrook, IL) in pediatric outpatient and in adult inpatient settings, where the assay was performed by non–laboratory-based physicians, and demonstrated more appropriate triage into isolation rooms.21, 22 Another study evaluated the FilmArray Respiratory Panel (BioFire Diagnostics, Inc., Salt Lake City, UT) when placed in a core laboratory of a regional children's hospital; 81% of influenza-positive patients received appropriate antiviral treatment within 3 hours of ED discharge, but the non–CLIA-waived test took a median time of 1.4 hours and was performed in the laboratory.²³ Modeling the use of rapid respiratory viral PCR in the ED suggested economic benefits but depends on the prevalence of the disease; treatment guided by physician diagnosis or rapid testing and treatment of all patients are more effective and less costly than no treatment.24, 25

To our knowledge, our study herein is the first to evaluate the impact of rapid influenza and RSV bedside PCR on immediate clinical management in a pediatric ED rather than through a post hoc survey. This approach capitalizes on the clinical judgment of physicians while they were still actively managing patients. The intended device is the Cobas Liat influenza A/B and RSV assays (Roche Molecular Diagnostics, Pleasanton, CA), which has a 20-minute turnaround time, sensitivity of 97.5% to 100%, and specificity of 100% in small studies.15, 26 The mPOCT is now FDA cleared and CLIA waived for nontechnician bedside testing of both influenza A/B and RSV (https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K153544; http://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfCLIA/Detail.cfm?ID=39763&NoClia=1, last accessed January 10, 2017); certifications were granted only after completion of this study. Furthermore, this bedside test has the same documented accuracy as the traditional laboratory-based respiratory PCR test.15, 18

The primary outcome for this study was reduction in ED LOS. The original ED LOS (triage to disposition) was compared with theoretical ED LOS (triage to mPOCT ordered, plus 20 minutes to allow for resulting of proposed test). Even allowing 20 minutes for results of this proposed bedside test, we still found that approximately 30 minutes less could have been saved had the rapid test been available for these patients. The 15% potential reduction in ED LOS could translate into markedly improved resource burden and could reduce ED crowding during peak respiratory illness season. Prior studies have considered 18 to 60 minutes to be significant reductions in ED LOS when assessing the value of interventions in the ED that monitor ED LOS as a primary or secondary outcome.11, 27, 28, 29, 30 Beyond charge savings for patients, hospitals may improve operational efficiency to see more new patients, curtail ambulance diversions, and avoid delayed care in time-sensitive diseases, although these metrics were not measured during this study. Given the long turnaround time in our laboratory comparator, substantial ED LOS improvement may be expected for any test that has a short turnaround time; however, tests that require longer than the proposed 33-minute improvement in ED LOS would be less likely to see improvement in this outcome because turnaround time may eclipse ED LOS improvement. Another possibility not explored through this study method would be the potential impact of initiating mPOCT testing at triage, which may further reduce ED LOS substantially, given the ED LOS among our patients exceeded 3 hours in most cases.

By supplanting blood work, UA, and CXR as a first-line diagnostic, a rapid viral test offers a fast and noninvasive evaluation of undifferentiated fever in children. For many patients with a positive viral result, physicians thought they would order no further ancillary testing, preventing unnecessary radiation exposure, phlebotomy, and urinary catheterizations in a young population. This is not surprising and is consistent with the published reports that show in young children aged 3 to 36 months who are at risk of severe bacterial infection the prevalence of severe bacterial infection is lower in febrile children with viral infections.³¹ Furthermore, Titus et al³² and Levine et al³³ independently demonstrated that, other than urinary tract infection, severe bacterial infections were substantially less likely to occur in young children (younger than 60 days or younger than 8 weeks) who tested positive for RSV. They found as a result there was reduced ancillary testing, resulting in substantial charge savings to patients, even when averaged across all patients rather than only patients with positive viral results.

ED-based satellite laboratories have been explored with some success as alternatives to central laboratories or bedside mPOCT devices throughout the ED.²⁷ When considering advantages of mPOCT compared with near-patient testing in an ED-based laboratory, bedside testing can avoid costs associated with maintaining an ED-based laboratory and hiring specialized technicians; mPOCT tests have been shown to be cost-effective compared with laboratory testing both in clinics and EDs.34, 35 However, mPOCT testing errors can occur, caused in part by high personnel turnover rates, lack of understanding about good laboratory practices, and inadequate training.36, 37 Before introducing any waived testing, clinical sites should follow recommendations developed by the CLIA Committee.³⁸

Previous work found that rapid testing did not significantly affect admission status¹⁴; most disposition destinations were unchanged regardless of outcome. In our study it appeared that physicians were more likely to change management and to rely on the results of rapid testing to do fewer tests in patients who looked well and could be considered for discharge. Although influenza and RSV are by far the most common viruses in our traditional PCR panel, only 30% of virus-positive patients tested positive for viruses other than RSV or influenza. Thus, it is unlikely that rapid testing alone has the breadth to inform isolation status decisions for admitting teams, given that this test cannot rule out other respiratory infections that would necessitate isolation on admission. Similarly, patients with sufficient acuity to require admission to the intensive care unit are unlikely to have disposition or ED evaluation markedly altered by rapid viral testing because criteria for admission to the intensive care unit usually depend more on physiological markers and disease presentation rather than specific viral causes.

Rapid viral testing previously has not been shown to improve antibiotic use in the ED.³⁹ We demonstrate a trend toward decreased antibiotic treatment among all patients, with a significant decrease in antibiotic prescription to discharged patients. We propose that this is because of the improved sensitivity and specificity of our proposed test compared with other antigen tests. This suggests that when viral testing is timely and accurate, it can play an important role in antimicrobial stewardship by identifying nonbacterial sources of fever. In contrast, antiviral usage increased, given rapid testing results, perhaps prompting an increase in likely effectiveness of treatment when viral presence has been confirmed versus relying on history and examination alone.

Clinical markers, history, and examination have repeatedly been shown to be poor predictors for presence of influenza, so results of influenza testing are highly relevant in the ED.6, 40, 41 In contrast to its recommendation against rapid antigen testing because of limited sensitivity, the Centers for Disease Control and Prevention suggests that PCR testing might be advantageous if the result would inform clinical decisions on use of antiviral treatment, antibiotic treatment, need for further diagnostic tests, consideration for home care, or recommendations for ill persons living with others who are at high risk of influenza complications.7, 42 Utility of this test is likely to change with different disease prevalence; for example, it is possible that significantly high local prevalence would obviate the need for confirmatory viral testing. A follow-up study using this device is being undertaken during the 2016 to 2017 ARI season.

Despite our unique study design to determine the potential clinical impact of the mPOCT by comparing a hypothetical rather than an actual test with our institutional standard, interpretation of our results are limited by the assumptions made. In this prospective real-time survey study, we made reasonable assumptions about assay time and accuracy based on known performance characteristics of the intended mPOCT. However, we recognize that our proposed turnaround time of 20 minutes is optimistic, given likely real-world delays depending on staff availability to run the bedside test, variable operator performance and reporting efficiency, or possible competition for single-use device. Time loss from such delays compared with centralized testing delays of transport and result notification would be better assessed with actual testing.

Nonetheless, our novel approach did allow us to directly compare patients with themselves, perfectly matching for severity of presentation and idiosyncrasies in patient management. Furthermore, to avoid any retrospective biases we solicited real-time physician decision making and opinions by asking them to assess the impact of mPOCT results on their current patient while they were actually managing the patient. This structure provides a framework for future investigations that aim to test the potential implementation and utilization of these new diagnostic tests, without the biases of post hoc questionnaires. For example, in this study we found the test to be useful in only cases involving active management and not for isolation status in the ED, and we recommend limiting the device to those patients. This is particularly applicable to emerging mPOCT platforms, given their rapid technologic evolution cycles, for which demonstration of clinical impact may hasten a lengthy test clearance process. Overall, we believe that the results are valid, but we also realize that an implementation study of the device in pediatric ED patients is needed to validate our results to ensure that physicians would actually follow through on their management plans.

Supplemental Data

Supplemental Figure S1.

Interview algorithm. Research assistants received immediate electronic notification when a respiratory panel was ordered. The attending physicians in the pediatric emergency department (ED) was called within 5 minutes of ordering the panel, and the above interview algorithm was followed to ascertain the intended real-time management plan, before patient disposition or availability of panel and ancillary testing results. RSV, respiratory syncytial virus.