Clinical Influenza Testing Practices in Hospitalized Children at United States Medical Centers, 2015-2018

Abstract

At nine US hospitals that enrolled children hospitalized with acute respiratory illness (ARI) during 2015-2016 through 2017-2018 influenza seasons, 50% of children with ARI received clinician-initiated testing for influenza and 35% of cases went undiagnosed due to lack of clinician-initiated testing. Marked heterogeneity in testing practice was observed across sites.

Millions of influenza cases and thousands of influenza-associated hospitalizations occur in children annually in the United States [1]. Clinical guidelines recommend influenza testing for all hospitalized children with an acute respiratory illness (ARI) while influenza is locally circulating [2]. Testing informs clinical decisions, infection control, and understanding of locally circulating pathogens, and although influenza antiviral therapy should be initiated immediately for suspected influenza in hospitalized patients to improve outcomes [3], it is rarely started without clinical testing [4].

The New Vaccine Surveillance Network (NVSN), a CDC-funded research network, performs protocol-driven influenza research testing among hospitalized children meeting criteria for ARI to determine influenza vaccine effectiveness annually [5, 6]. The network also collects details on independently performed clinician-initiated influenza testing, allowing us to examine patterns of clinician-driven testing in patients hospitalized with ARI. Within NVSN we: (1) evaluated the frequency of influenza testing in children with ARI and how clinician-initiated testing changed over three seasons; (2) assessed the heterogeneity of testing practices by age and hospital site; and (3) evaluated the missed burden of hospitalized influenza cases due to lack of clinical testing. We hypothesized that a large proportion of hospitalized children with ARI do not receive influenza testing, and this might result in a substantial burden of missed cases.

METHODS

NVSN enrolled children (<18 years) at eight academic medical centers (9 individual hospitals). One hospital participated only during the 2015-2016 influenza season and one only during 2016-2017 and 2017-2018 seasons. Children were eligible if they resided within the surveillance area and visited the emergency department (ED) or were admitted to the hospital within 48 hours of enrollment and had ARI or a related syndrome (eg, asthma exacerbation or pneumonia) in 2015-2016 [6]. During 2016-2017 and 2017-2018 seasons, the same eligibility criteria were used, although an additional symptom criterion was added (≥1 symptom of ARI without a known non-respiratory cause) [5]. Children were not eligible if they were seen in the ED but not admitted, symptom onset was >7 days before presentation, or they received influenza antiviral therapy prior to admission.

Data including clinician-initiated influenza testing were collected through medical records review of enrolled patients by trained study staff. Staff collected data on influenza testing performed either by rapid test or a molecular respiratory virus panel. For rapid tests, information on assay type (antigen vs molecular) was specified only for 2017-2018. We considered a patient clinically tested for influenza if ≥1 assay was performed throughout the hospitalization.

We calculated the number and percentage of children with ARI who received clinician-initiated testing by age (0-4 and 5-17 years) between December and April. Because research definitions for ARI may lack specificity and because of minor differences in research ARI criteria used between seasons, we also assessed clinician-initiated testing restricted to participants with ARI who also had one or more International Classification of Disease, Clinical Modification, 10th Edition (ICD-10-CM) codes for acute upper respiratory infection (J00-J06), influenza and pneumonia (J09-J18), or other acute lower respiratory infections (J20, J22) (Supplementary Appendix A, Supplementary Table 1) [7]. We excluded bronchiolitis (J21) as guidelines recommend laboratory testing not be routinely obtained [8]. Trends in the proportion tested over three influenza seasons were assessed using a linear mixed model treating site as a random effect and weighted for the number of enrolled participants. We further evaluated testing by hospital site, assessing heterogeneity using chi-square tests.

To provide insight into the missed burden of hospitalized influenza illness in children with ARI, we determined the proportion who tested positive for influenza by research RT-PCR testing by clinician-initiated testing status. Assuming research RT-PCR testing had near-perfect specificity, we assessed the number and proportion of children with influenza diagnosed through research testing who were not diagnosed by providers due to lack of clinician-initiated testing. Local clinicians were not aware of research testing results. To compare testing practices with children, we also evaluated clinician-initiated influenza testing practices among adults with ARI enrolled in a separate hospital-based adult vaccine effectiveness network (methods and results in Supplementary Appendix B). P-values <.05 were considered significant. Analyses were performed using SAS Version 9.4 (Cary, NC, USA) and R Version 3.6.1 (Vienna, Austria).

RESULTS

During three seasons, 7356 children at 9 hospitals in 8 states with ARI were included. Fifty percent (3688/7356) of included children received clinician-initiated testing for influenza (Table 1). Among patients who were clinically tested, 76% (2802/3688) were tested using a molecular respiratory virus panel. Restricted to those with ARI plus relevant ARI-related ICD-10-CM codes, 56% (1552/2777) were clinically tested for influenza. In comparison, among adults with ARI clinical testing was performed in 64% overall and 78% enrolled with ARI who also had an included ICD-10-CM diagnostic code (Supplementary Appendix B).

Table 1.

Clinician-Initiated Influenza Testing by Season and Age Group for Patients Meeting Criteria for Acute Respiratory Illness (ARI) or with Pneumonia and Influenza (P&I) or ARI International Classification of Disease, 10th Edition (ICD-10-CM) Discharge Diagnoses

Population and Test Type	2015-2016 Season	2016-2017 Season	2017-2018 Season
0- to 4-year-olds
Acute respiratory illnessa	n = 2025	n = 1923	n = 1860
Any influenza testing, % (No.)	52% (1050)	48% (927)	53% (989)
Molecular respiratory panel, % (No.)	38% (772)	40% (760)	39% (726)
Singleplex/rapid molecular, % (No.)	—	—	6% (109)
Rapid antigen, % (No.)b	16% (330)	14% (277)	16% (292)
ICD-10-CM discharge diagnosis	n = 653	n = 639	n = 671
Any influenza testing, % (No.)	56% (364)	54% (347)	56% (378)
Molecular respiratory panel, % (No.)	39% (257)	42% (271)	39% (265)
Singleplex/rapid molecular, % (No.)	—	—	7% (46)
Rapid antigen, % (No.)b	19% (125)	20% (129)	19% (128)
5- to 17-year-olds
Acute respiratory illnessa	n = 525	n = 549	n = 474
Any influenza testing, % (No.)	38% (202)	49% (268)	53% (252)
Molecular respiratory panel, % (No.)	30% (158)	39% (215)	36% (171)
Singleplex/rapid molecular, % (No.)	—	—	4% (21)
Rapid antigen, % (No.)b	13% (66)	17% (91)	20% (93)
ICD-10-CM discharge diagnosis	n = 274	n = 278	n = 262
Any influenza testing, % (No.)	51% (139)	58% (161)	62% (163)
Molecular respiratory panel, % (No.)	40% (110)	44% (122)	39% (103)
Singleplex/rapid molecular, % (No.)	—	—	6% (16)
Rapid antigen, % (No.)b	18% (50)	24% (68)	26% (68)

No. = number tested.

Season for defined here as December 1 through April 30.

^aNumbers of individual test types may add up to greater than total number tested as some patients were tested using more than one type of assay.

^bThe New Vaccine Surveillance Network only specified if the rapid test was antigen-based or molecular during the 2017-2018 season. Among patients with acute respiratory illness during the 2017-2018 season, 130/515 (25%) of rapid tests were molecular with the rest rapid antigen tests. For the 2015-2016 and 2016-2017 seasons, we classify all rapid tests in the table as antigen-based assays in the table.

We did not observe a temporal trend in clinician-initiated testing across seasons (P = .20) for children with ARI; however, we did find clinical testing significantly increased across the seasons (P = .019) for children with ARI plus an acute respiratory infection-associated ICD-10-CM diagnosis, from 54% in 2015-2016 to 58% in 2017-2018 (Figure 1). Little change in testing was observed in younger children (aged 0-4 years), including those with ARI as well as those with ICD-10-CM diagnostic codes for respiratory infections. Significant heterogeneity was observed in testing by hospital (all P < .001; Supplementary Table 2). Among children with ARI, hospital testing ranged by as much as 27% (76/278) to 88% (188/214), during the 2016-2017 season. Among the 50% of children enrolled with ARI who received clinician-initiated testing, 8% (294/3688) had a positive research influenza RT-PCR test result. Among enrolled children who did not receive clinician-initiated testing, 4% (155/3668) of research RT-PCR tests were positive. Thus, assuming perfect specificity of the research RT-PCR, 35% (155/449) of the total burden of influenza was missed due to the absence of clinician-initiated testing.

Figure 1.

Percentage of children with acute respiratory illness (ARI) receiving influenza clinician-initiated testing by influenza season and hospital in children* enrolled in the New Vaccine Surveillance Network—the United States, 2015-2018. *Sites with <50 patients enrolled with ARI during a given season are not shown in the figure. Lines are used to connect individual hospital sites values by season. Some hospitals did not participate during all seasons or may have had <50 patients included with ARI during the season.

Discussion

Results from this study conducted over three recent influenza seasons in the United States provide insights into clinician-initiated influenza testing among hospitalized children with ARI. The proportion tested for influenza was relatively low at 50% but differed markedly by hospital. Restricted to children with ARI plus a respiratory illness-related ICD-10-CM code, performance of clinical testing increased to 56%. A lack of clinician-initiated testing resulted in a failure to identify about one-third of children with influenza detected by research testing. These findings suggest there are some opportunities to improve decision pathways for influenza testing.

For children hospitalized with ARI, decisions about whether to test for influenza or other respiratory viruses may involve a variety of considerations. Indiscriminate testing for circulating respiratory viruses can lead to added laboratory costs and low diagnostic yield; even among children clinically tested for influenza at NVSN sites, research RT-PCR testing was positive in <10%. However, the use of guideline-recommended testing of hospitalized children with ARI also has a variety of direct and indirect benefits including promoting use of influenza antiviral therapy, avoiding antibiotic overuse, hospital cohorting for infection prevention, and understanding of the epidemiology of locally circulating respiratory pathogens. The marked heterogeneity observed in testing practices between hospitals suggests under-testing in some hospitals based on local practices. Future analyses should evaluate factors associated with influenza testing practices in hospitalized children and optimizing decision pathways.

In addition to supporting clinical decision making, findings from this study support public health efforts to assess the true burden of severe illness due to influenza [9]. Surveillance systems that rely on clinician-testing alone will systematically underestimate the burden of respiratory infections. Accurate estimates of the burden of respiratory infections—both diagnosed and undiagnosed—have broad implications. These include promoting public awareness of the impact of infections and the importance of receiving vaccination, promoting non-pharmaceutical interventions, and guiding health policy decisions.

One strength of our analysis is that we examined testing practices among hospitalized patients using 2 criteria. This included protocol-defined ARI criteria which are meant to be sensitive but lack specificity. We also evaluated clinician-initiated influenza testing among patients with respiratory illness-related ICD-10-CM codes which are more specific and capture illnesses assigned to the patient during admission. Although clinical testing was higher in children with relevant ICD-10-CM diagnoses, among those with included ICD-10-CM codes only 56% received clinician-initiated testing over three influenza seasons.

Our study has several limitations. First, the pre-specified ARI symptoms were intentionally broad for research purposes; however, case definitions are less specific for influenza and could lead to an underestimate of the proportion clinically tested. Secondly, we collected data during only three influenza seasons and clinical testing practices by be influenced by additional time-varying factors such as circulation of SARS-CoV-2. Finally, we captured data from hospitals that are geographically diverse but may not be generalizable.

CONCLUSION

Clinician-initiated influenza testing among children hospitalized with ARIs occurred in one-half of patients and was similar across three seasons. Opportunities exist for reinforcing influenza testing recommendations for patients hospitalized with ARI.

Notes

Acknowledgments. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. We would like to thank Lindsay Kim MD, MPH, Brian Rha MD, Constance Ogokeh MPH, Joana Lively MPH, and Lauren Grant MA at CDC for their assistance with the project.

Financial support. This work was supported by the Centers for Disease Control and Prevention (CDC) [5U01IP001035-02] and, at the University of Pittsburgh, by the National Institutes of Health (NIH) [UL1TR001857]. Vanderbilt University Medical Center also received support from the National Center for Advancing Translational Sciences Clinical Translational Science Award Program (5UL1TR002243-03).

Potential conflicts of interest. Dr J. A. B. reports grants from CDC, during the conduct of the study; Dr J. A. E. reports personal fees from Sanofi Pasteur, personal fees from Meissa Vaccines, grants from AstraZeneca, grants from Novavax, grants from GlaxoSmithKline, grants from Merck, grants from Gates Ventures, outside the submitted work; Dr M. G. reports grants from CDC, during the conduct of the study and grants from CDC, grants from CDC-Abt Associates, grants from MedImmune/AstraZeneca, outside the submitted work; Dr N. B. H. reports grants from CDC, during the conduct of the study and grants from Sanofi, grants from Quidel, personal fees from Genentech, outside the submitted work; Dr C. J. H. reports grants from DAIDS, Pfizer, Merck, and GlaxoSmithKline outside the submitted work and from CDC during the conduct of the study; Dr E. T. M. reports grants from Centers for Disease Control and Prevention, during the conduct of the study and personal fees from Pfizer, grants from Merck, outside the submitted work; Dr D. B. M. reports personal fees from Seqirus, grants and personal fees from Pfizer, personal fees from Sanofi Pasteur, personal fees from GlaxoSmithKline, personal fees from Dynavax, outside the submitted work; Dr A. S. M. reports personal fees from Sanofi, personal fees from Seqirus, personal fees from Roche, outside the submitted work; Dr L. C. S. reports grants from CDC, during the conduct of the study; Dr R. S. reports grants from Cepheid, grants from BioFire, grants from Becton and Dickinson, grants from Abbott diagnostics, grants from Hologic, outside the submitted work; Dr F. P. S. reports grants from CDC, during the conduct of the study, grants from Ansun, outside the submitted work, and personal fees from Janssen, outside of the submitted work; Dr L. S. S. reports grants from CDC, during the conduct of the study; Dr G. A. W. reports grants from US DHHS/CDC, during the conduct of the study and personal fees from Merck & Company, personal fees from ReViral, outside the submitted work; Dr J. V. W. reports grants from CDC, during the conduct of the study and personal fees from Quidel, personal fees from ID Connect, personal fees from GlaxoSmithKline, outside the submitted work; Dr R. K. Z. reports grants from CDC, during the conduct of the study and grants from Sanofi Pasteur, outside the submitted work; all other authors have reported no potential conflicts of interest to disclose. All authors have submitted the ICMJE Form for Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.