Antiviral Use in Canadian Children Hospitalized for Influenza

The study population comprised patients aged 0–16 years ascertained through active surveillance for laboratory confirmed influenza admissions from September 1, 2010, to June 30, 2019, at the 12 tertiary care pediatric hospitals participating in the Canadian Immunization Monitoring Program, ACTive (IMPACT).²⁹  These centers admit >75 000 children annually, account for ∼90% of pediatric tertiary care beds in the country, receive referrals from all provinces and territories, and serve a population of ∼50% of Canada’s children.³⁰  During the study period, all centers tested children admitted with acute respiratory infection for influenza, guided by their local policies and test availability. Patients with hospital LoS <1 day, receipt of antiviral agents before admission to the IMPACT hospital, and nosocomial cases (symptom onset ≥72 hours after admission) were excluded from this study. Institutional ethics approval was obtained at each center.

Trained nurse monitors prospectively screened daily laboratory results and admission lists for eligible cases. Audits of hospital discharge abstract databases were performed yearly to ensure case ascertainment completeness. Data were abstracted from medical charts.

The primary outcome was receipt of anti-influenza antiviral treatment during hospitalization. The secondary outcome was timing of antiviral receipt during admission. Exposure variables included demographics, health status (by chronic conditions known to be risk factors for complicated infection³¹ ), influenza season (eg, 2010–2011), admitting center, availability of local influenza treatment guidelines during that season at the admitting center, timing of admission within influenza season (screening for influenza at IMPACT centers occurs year-round, and we defined “peak” season as the period during which the national influenza test positivity proportion was at least 15%, as reported by the Public Health Agency of Canada⁷ ), test method used for diagnosis, timing of availability of laboratory confirmation of influenza infection in relation to admission, influenza vaccination status for that season, antibiotic prescription, and measures of illness severity (mortality, ICU admission, respiratory support [continuous positive airway pressure (CPAP), bilevel positive airway pressure (BiPAP), conventional or high-frequency ventilation, and extracorporeal membrane oxygenation (ECMO)], ICU, and hospital LoS).

Data were analyzed by using R, version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was assessed by using 2-tailed tests, with an α of 0.05. Descriptive statistics were calculated, medians of continuous variables were compared across strata using the Wilcoxon rank test, and proportions were compared by using the Pearson’s χ² test. The Mann-Kendall test was used to analyze temporal trends in antiviral use. Correction for multiple hypothesis testing was not performed. Logistic regression analyses were used to determine crude and adjusted odds ratios (aORs) and corresponding 95% confidence intervals (CIs) for factors associated with antiviral use, and for factors associated with early (during first 2 days of admission) versus late (after first 2 days of admission) initiation of antiviral agents. Given the substantial proportion of missing data for ethnicity (47%) and vaccination status (16.8%), no inferential statistics were performed on these variables. For the multivariable models, we considered variables that were statistically significant on univariable analyses, and, a priori, included plausible confounders based on literature or clinical experience. Model selection was guided by using the Akaike Information Criteria, using backward selection.

During the study period, 7946 laboratory-confirmed influenza admissions were recorded. Among these, 276 had a hospital LoS <1 day, 103 had received antiviral agents before hospitalization at the IMPACT center, and for 22 cases, data regarding receipt of antiviral agents were unknown. After these exclusions, we included 7545 cases (Fig 1).

Median patient age was 3.0 years (interquartile range [IQR]: 1.1–6.3) and 57.4% were male (Table 1). Median number of cases admitted per influenza season was 680 (range 561–1298). The total number of cases at each IMPACT center during the study ranged from 185 to 1132. Eight IMPACT centers (75%) had local influenza treatment guidelines (adapted from national guidance documents), and onset of guideline availability ranged from before 2010–2011 to 2016–2017. Among patients ≥6 months of age with known vaccination status, only 690 of 5402 (12.8%) had documented receipt of influenza vaccine that season. Influenza A virus was detected in 5345 cases (70.8%), and influenza B was detected in 2165 (28.6%). Among the 1926 type A isolates subtyped, 50.3% were A/H3N2 and 45.9% A/H1N1 2009.

The median duration of symptoms before presentation was 3 days (IQR: 2–5). At least 1 chronic health condition was found in 3434 (45.4%) children. Seventy-two percent received antibiotics; laboratory confirmed bacterial infections (positive tissue, aspirate, blood, cerebrospinal fluid, and/or urine culture) were noted in only 508 (6.7%) patients. ICU admission occurred in 1252 (16.5%) cases, and 795 (63.5%) of these required respiratory support. Forty-two patients (0.5%) died from influenza-related complications.

Overall, 3122 patients (41.3%) were prescribed antiviral agents; 3118 (99.9%) received oseltamivir. Antiviral use increased from 19.9% in the 2010–2011 season to 59.6% in 2018–2019 (P for trend = .001) (Fig 2). Prescribing varied widely across sites (range 10.2% to 81.1%). Almost all cases (93%) had influenza test results available within 48 hours of admission. Among patients who received antiviral agents, 2551 (81.7%) received them within 2 days of admission (Supplemental Fig 5). The proportion treated decreased with time from symptom onset to admission (<2 days, 48.3%; 2–4 days, 44%; >4 days, 35%). Antiviral use increased with age, from 32% in those 0 to 5 months old to 48.6% in those >5 years old. Children with ≥1 chronic health condition were more likely to receive antiviral agents (52.7% vs 36.7%; P < .001); this was also seen across most individual medical conditions (Table 2). Median time to treatment after admission did not differ significantly between those with or without an underlying chronic health condition (2 ± 1.32 vs 2 ± 1.02 days; P = .06). Antiviral prescribing increased similarly over time for patients with or without a chronic health condition (Fig 3) and across age groups (Fig 4). Prescription of antibiotics was more frequent in children who received antiviral agents (77.3% vs 69.6%; P < .001). Patients who received antiviral agents had increased markers of disease severity (median hospital LoS, 4 vs 2 days, P < .001; ICU admission, 27.8% vs 8.68%, P < .001; median ICU LoS, 3 vs 2 days, P < .001; influenza-related mortality, 0.9% vs 0.2%, P < .001).

Multivariable logistic regression identified increased odds of receipt of antiviral therapy with advancing age (aOR 1.04 [95% CI, 1.02–1.05]), more recent season (highest aOR 9.18 [6.70–12.57] for 2018–2019), admission during peak season (aOR 1.37 [1.19–1.58]), local treatment guideline availability during that season at that center (aOR 1.54 [1.17–2.02]), timing of availability of laboratory confirmation of infection (highest aOR 2.67 [1.97–3.61] for result availability before admission), radiographic pneumonia (aOR 1.39 [1.20–1.60]), coreceipt of antibiotic therapy (aOR 1.51 [1.30–1.76]), ICU admission (aOR 3.62 [2.88–4.56]), and respiratory support (aOR 1.57 [1.19–2.08]). Among chronic health conditions, cancer was most strongly associated with antiviral therapy, (aOR 4.81 [3.61–6.40]). Influenza B virus infection demonstrated lower odds (aOR 0.81 [0.70–0.94]) of receiving antiviral therapy. In a model restricted to children that received antiviral agents, early (during first 2 days of admission) initiation of treatment was strongly associated with timing of availability of laboratory confirmation of infection and inconsistently associated with more recent seasons (Supplemental Table 3).

North American influenza clinical practice guidelines, including those of the Infectious Diseases Society of America,¹⁸  American Academy of Pediatrics,³²  and the Canadian Pediatric Society,³³  recommend antiviral treatment of all children hospitalized with influenza. Despite this, our country-wide hospital-based active surveillance study comprising >7500 pediatric admissions for laboratory confirmed influenza during 9 postpandemic seasons found that the overall use of antiviral agents was only 41%. Prescribing increased over time, demonstrated wide variation across participating centers, and was associated with the timing of admission relative to peak influenza circulation, timing of availability of influenza test results, and availability of local influenza treatment guidelines. Furthermore, we identified patient-level factors associated with antiviral use which included older age, presence of underlying chronic health conditions that are risk factors for severe illness, infecting virus type, radiographic pneumonia, coreceipt of antibiotics, and need for intensive care and respiratory support.

The use of neuraminidase inhibitors in adults and children in randomized placebo-controlled trials of the treatment of influenza in outpatients has been demonstrated to reduce duration of illness and viral shedding, and risk of influenza-associated complications.^12,34  The evidence base for antiviral treatment of influenza in the hospital setting is weaker. The only published trial in which researchers evaluated neuraminidase inhibitors among children admitted with influenza was terminated early with only 21% of the targeted population enrolled.¹⁶  However, mounting evidence from observational studies in hospitalized children and adults suggests clinical benefit (including decreased LoS, health care costs, ICU admission, and mortality) of treatment with neuraminidase inhibitors, especially in high-risk populations and when initiated early (within 48 hours of illness onset or of hospitalization).^21,23,35  Despite this, our study and others show that a substantial number of hospitalized children with influenza do not receive antiviral treatment. In a retrospective cohort of ∼36 000 children hospitalized with influenza during 2007–2015 in the United States, 69% received antiviral treatment; 30% of children at high risk were not treated.²⁷  Among influenza hospitalizations (adults and children) in Ontario from 2004–2005 to 2013–2014, the percentage treated increased from 29% before the 2009 pandemic to 74% during the pandemic, and 55% to 65% thereafter.³⁶  Similarly, we found that use rates increased over time, climbing to almost 60% during 2018–2019; this trend was observed across centers and patient subpopulations. Antiviral use among children most likely to benefit from treatment, those with chronic health conditions, more than doubled over the study period to nearly 70% in 2018–2019.

Guidelines recommend initiation of antiviral treatment within 48 hours of onset of illness in outpatients, because optimal benefits are obtained with early treatment.^17,18,37  In this study, only a quarter of children presented within this time interval; delayed presentation (ie, >48 hours) may have led some clinicians not to use antiviral agents. However, antiviral agents are recommended for all patients hospitalized for influenza, even if treatment is started >48 hours after onset of illness.^17,18,37  Like others, we observed that likelihood of antiviral use increased with rapidity of influenza testing³⁸ ; moreover, prompt availability of test results was strongly associated with earlier initiation of antiviral therapy. These findings suggest that integrating novel rapid and accurate molecular assays to emergency department diagnostic algorithms, especially for at-risk children being hospitalized, would facilitate early diagnosis and treatment.³⁹  Furthermore, although >80% of children treated with antiviral agents received them within 48 hours of admission, only 42% received them on the day of admission. Given the benefits of earlier treatment, antiviral agents should be initiated empirically for children hospitalized with suspected influenza and not delayed while awaiting the results of laboratory testing. Additionally, considering that only a quarter of patients in our study presented to an IMPACT hospital within 48 hours of symptom onset, primary care clinicians of at-risk children should counsel parents to seek care early if their child develops influenza-like symptoms.

We observed wide variation of antiviral use across IMPACT centers (range, 10.2% to 81.1%). Similarly, Stockmann et al²⁷  found that antiviral use ranged from 42% to 90% across 46 freestanding US children’s hospitals. Although most IMPACT centers had local antiviral guidelines, such guidelines were not always in place during earlier seasons. We found that the availability of local treatment guidelines was significantly associated with antiviral use, strengthening the call for local treatment protocols to improve adherence to national recommendations. Furthermore, quality improvement methods such as awareness modules, biweekly flyers, failure tracking, and similar approaches have also been shown to increase antiviral use among children hospitalized with influenza.⁴⁰ 

Additional hospital-level factors not evaluated in this study, such as the population served, referral patterns, and the presence of an antimicrobial stewardship program, may also influence local prescribing cultures. Moreover, recommendations for treatment of children hospitalized with influenza are based only on observational data. Randomized clinical trials of neuraminidase inhibitors among outpatients have shown modest benefits but with accompanying risk of adverse events, primarily gastro-intestinal side-effects, which may be transient and mitigated when oral oseltamivir is taken along with food.⁴¹  However, the perception of a questionable risk/benefit profile may lead to hesitancy to prescribe antiviral agents among some clinicians.^27,42 

It is concerning that antibiotics were used almost twice as frequently as antiviral agents in this cohort of children admitted with laboratory-confirmed influenza. Although the IMPACT data set is limited in its ability to provide antibiotic treatment details (spectrum of activity; timing and duration of use; complete versus incomplete courses), only 6.7% of children had a laboratory-confirmed concomitant or secondary bacterial infection and we presume that most physicians were empirically treating pneumonia and/or acute otitis media, which rarely have a laboratory-confirmed etiology. Our findings are similar to those from other studies evaluating antimicrobial use in hospitalized influenza cases^36,43  and highlight the difficulty clinicians face distinguishing bacterial versus viral causes of severe lower respiratory tract infections because of overlap in presentations, risk of bacterial superinfection, and challenges in obtaining samples from the lower respiratory tract for microbiologic testing.^44,45  Multifaceted interventions using combinations of education, rapid viral testing, biomarkers, and audit with feedback may be needed to reduce antibiotic overuse in influenza-associated hospitalizations.^46–48 

We found that older age was associated with antiviral use. More “classic” presentation among older children,⁴⁹  lack of approval of oseltamivir use by Health Canada in infants <12 months of age, and evolving guidance on oseltamivir use in younger children⁴³  could be reasons for antiviral use increasing with older age. Reports of neurotoxicity from animal models when oseltamivir was used in the infant rats⁵⁰  had raised concerns for its use in young children; however, no such neurotoxicity has been observed in human studies.¹⁷  Neuraminidase inhibitors are not currently authorized in Canada for the treatment of seasonal influenza in infants aged younger than 1 year, and oseltamivir use in infants is recommended on a case-by-case basis, based on illness severity.¹⁷ 

The presence of chronic conditions known to be risk factors for complicated infection were significantly associated with antiviral use in a multivariable logistic regression model. Children with these comorbidities may present more severe disease, which may partly explain greater antiviral use in this population; nevertheless, after adjustment for requirement for ICU and respiratory support, these conditions remained associated with antiviral prescription. This suggests that clinicians recognize that these children are at greater risk of severe outcomes and potentially benefit most from antiviral treatment. However, earlier treatment was not observed in patients with comorbidities, reinforcing the need for rapid diagnostics or prompt empirical therapy while awaiting influenza test results for at-risk children.

Our study is limited by its retrospective design. Although case reporting was conducted prospectively by nurse monitors using a standardized reporting form, data were collected from the medical chart and could not capture clinical decision-making processes. Data regarding the receipt of influenza vaccine and ethnicity were frequently missing. Furthermore, although site investigators are questioned annually on testing practices, we cannot ascertain whether all children admitted with acute respiratory infection received influenza tests. Clinical data were limited regarding the timing of appearance of specific signs or symptoms during the course of illness. Moreover, although we were not able to include or control for measures of severity at presentation or admission to hospital, we attempted to overcome this limitation by including need for ICU admission and respiratory support as proxies for disease severity in our multivariable model. Finally, the external validity of this study is limited by the fact that IMPACT conducts surveillance in tertiary care centers in Canada; the management of pediatric influenza associated admissions in community hospitals may be different. Despite these limitations, our study is strengthened by its representation of a nationwide active surveillance cohort of children hospitalized with laboratory confirmed influenza admitted over almost a decade, and it is one of the largest such studies in the postpandemic period.

Antiviral medications are underused among children hospitalized for influenza in Canadian pediatric hospitals. Moreover, nearly half of children with chronic health conditions placing them at risk for severe outcomes were not treated. However, an increase in use overall, and in children at high risk, was observed over time. Despite national clinical practice guidelines, there was wide variation in antiviral prescribing practices across the country, and a high rate of antibiotic use was noted. We also identified patient and hospital-level characteristics independently associated with antiviral prescribing. Prompt availability of influenza test results was strongly associated with earlier initiation of antiviral therapy. Taken together, these findings call for multifaceted interventions to strengthen adherence to local and national influenza treatment guidelines and antimicrobial stewardship practices.

We acknowledge the IMPACT nurse monitors, Annick Audet (IMPACT Nurse Monitor Liaison), the staff of the data center, including Kim Marty (Data Manager) and Jennifer Mark (Data Research Assistant), and Melanie Laffin Thibodeau (Manager, Surveillance, Canadian Paediatric Society).

Investigators and centers participating in this IMPACT project included: Natalie Bridger, Cheryl Foo Janeway Children’s Health & Rehabilitation Centre, St. John’s, NL; Scott A. Halperin, Karina Top, IWK Health Centre, Halifax, NS; Roseline Thibeault, Centre Mere-Enfant de Quebec, CHUL, Quebec City, QC; Dorothy L. Moore, Marie-Astrid Lefebvre, Jesse Papenburg, The Montreal Children’s Hospital, Montreal, QC; Marc Lebel, Centre hospitalier universitaire Sainte-Justine, Montreal, QC; Nicole Le Saux, Children’s Hospital of Eastern Ontario, Ottawa, ON; Dat Tran, Shaun K. Morris, The Hospital for Sick Children, Toronto, ON; Joanne Embree, Winnipeg Children’s Hospital, Winnipeg, MB; Ben Tan, Athena McConnell, Rupeena Purewal, Royal University Hospital, Saskatoon, SK; Taj Jadavji, Cora Constantinescu, Alberta Children’s Hospital, Calgary, AB; Wendy Vaudry, Catherine Burton, Stollery Children’s Hospital, Edmonton, AB; Julie Bettinger, Laura Sauvé, Manish Sadarangani, BC Children’s Hospital, Vancouver, BC.

We thank Chelsea Caya and Ian Schiller (Research Institute of the McGill University Health Centre) for helpful feedback on statistical programming.

AIC

Akaike Information Criteria

aOR

adjusted odds ratio

BiPAP

bilevel positive airway pressure

CI

confidence interval

CPAP

continuous positive airway pressure

ECMO

extracorporeal membrane oxygenation

IMPACT

Canadian Immunization Monitoring Program, ACTive

IQR

interquartile range

LoS

length of stay