Background

International rates of hospitalization due to acute upper and lower respiratory disease are highest among infants aged <1 year old [1–3]. Influenza virus [1, 2, 4–6] and respiratory syncytial virus (RSV) [7, 8] infections are recognized as leading contributors to this burden; yet, existing studies typically underestimate the frequency of severe influenza and RSV diseases that require hospitalization among infants for multiple reasons. First, clinical complications from influenza virus infections, such as pneumonia and bronchiolitis, often occur days after the primary infection when molecular diagnostics may be less sensitive to detect reduced virus shedding by the time of hospitalization.Footnote 1 However, as demonstrated by a study in Thailand, serologic diagnostics may be able to detect antibody seroconversion following influenza infection even among hospitalized pneumonia patients who test negative for influenza using molecular diagnostics [9–11]. Likewise, serology may be a complementary tool to molecular diagnostics in determining recent RSV infections [12, 13].

Second, existing surveillance platforms often overlook non-febrile infections and non-respiratory manifestations of influenza and RSV infection [14–17]. For example, in a study of children aged <3 years presenting with fever to emergency departments in France, about 1 in 3 children with laboratory-confirmed influenza presented with a non-respiratory disease (including febrile seizures, sepsis-like syndrome, and gastroenteritis) [14]. Likewise, RSV is typically considered an etiologic agent in respiratory tract infections during the RSV season [18, 19], but it is also an important cause of apnea in preterm infants [20–22] and can present with shock and as a sepsis-like syndrome in young infants [23–25]. Additionally, both influenza and RSV infections often presents without fever in infants [7, 17, 26]. Thus, the true burden and clinical manifestations of both influenza and RSV among infants might be under-represented if subjects are selected for pathogen testing among those with only respiratory symptoms or only fever. Finally, confidence in the etiologic role of influenza and RSV in infant hospitalization requires comparative information on the prevalence of these infections among non-ill infants at the same time and from the same communities, which is often overlooked in epidemiologic studies of this population [27].

This U.S. Centers for Disease Control and Prevention (US CDC) funded multi-site prospective study will enroll approximately 3,000 infants aged <1 year hospitalized for any acute illness and a comparison sample of approximately 1,400 non-ill infants.Footnote 2 Here we provide an overview of the design and methods of the Influenza and RSV in Infants Study (IRIS) currently starting year 2 of a 2- to 3-year study in Albania, Jordan, Nicaragua, and the Philippines. Our study aims are to assess the frequency of influenza- and RSV-associated hospitalizations (i.e., severe disease) for respiratory and non-respiratory diseases, describe the clinical features of these infections and the predictors of intensive care (i.e., very severe disease), and describe the antibody response to influenza virus and RSV infections. Table 1 lists the knowledge gaps we identified within each of these aims and the study features intended to address these gaps.

Table 1 Study goals and features intended to address specific knowledge gaps
Full size table

Methods/Design

This prospective study consists of a hospital-based cohort and a cross-sectional sample of non-ill infants enrolled over 3 years in hospitals and ambulatory care settings serving infants in four countries (Table 2).Footnote 3 All study sites use a common protocol and data collection tools (using personal computers or mobile devices).Footnote 4 The design and screening methodologies for the hospital and community studies are described separately (below) followed by a description of common laboratory, statistical, and data management procedures.

Table 2 Network study countries, sponsors, and enrollment sites
Full size table

Hospitalized infants study design

For each study year, each site is enrolling approximately 350–500 infants aged <1 year old hospitalized with an acute medical illness during periods of local influenza circulation in year 1 and during periods of either influenza viruses or RSV circulation starting in year 2 (and for any subsequent years of the study), as indicated by local and national virus surveillance platforms.Footnote 5 We intend to begin enrollment during the early phase of local influenza virus and RSV circulation and continue past the peak through the decline in local circulation if possible. Enrollment goals are set per week in order to sample infants and extend the use of resources across the study period; enrollment can potentially expand during periods of peak circulation of either influenza or RSV.Footnote 6 Each day, study staff use hospital admission records (including presenting complaints and preliminary admission diagnoses) to register all acutely ill infants and stratify them into respiratory versus non-respiratory illness groups and general ward versus intensive care unit (ICU) placements (Fig. 1).Footnote 7 Infants within the resulting four strata are randomly sorted and prioritized in order to approach a mixture of patient types until the maximum daily enrollment in each group has been met. Footnote 8 Infants are screened for eligibility within 24 h of hospital admission to confirm that the family lives in the local area and that the illness onset was within the prior 10 days. Only basic descriptive information is then recorded for infants who are ineligible, are not approached, or whose parents refuse informed consent; these data will be used to assess possible selection bias and to describe the denominator of weekly hospitalized acutely ill infants. Study staff collect other hospital census information daily, including the number of available pediatric general ward and ICU beds, to consider differences in access to hospital care over time and between study sites.

Fig. 1
figure 1

Steps in Hospital Enrollment and Follow-up

Full size image

After completing written informed consent, extensive socio-demographic [28, 29], epidemiological, and clinical information is collected at enrollment, at a 3–5 week follow-up encounter, and through hospital record abstraction (summarized in Table 3). Respiratory specimens using combined nasal and oropharyngeal (OP) swabs (or endotracheal aspirate for intubated infants) and up to 2.5 ml of blood are collected by trained staff within 24 h of admission.Footnote 9 Information on socio-demographic characteristics, medical history, maternal and infant vaccination history (documented from vaccination cards when available), and symptoms of the acute illness are gathered from parents (or guardians).

Table 3 Key Variables and Sources of Information for Hospitalized and Non-Ill Infants
Full size table

Hospital records are abstracted to document clinical signs and measurements at admission, including clinician-measured temperature, respiration rate, and oxygen saturation. For each subsequent day of hospitalization, we document clinical signs and the lowest and highest measurements of vital signs and oxygen saturation. Clinical diagnoses at admission and discharge are recorded, including International Classification of Diseases (ICD) codes for discharge diagnoses. Clinical interventions, such as oxygen support, and treatments (including antibacterial and antiviral agents) are documented throughout hospitalization. Clinical laboratory values (such as white blood cell count) and test results (including bacterial, viral, and parasite detection), chest radiographic information (including digital images, if available), and other clinical assessments carried out as part of regular clinical care are also recorded. We also document hospital readmissions within 30 days of discharge for the index illness.

All infants in the hospital-based cohort complete a follow-up assessment and a convalescent blood draw (up to 2.5 ml of blood) 3–5 weeks after enrollment through hospital or home visits. Information on the infant’s health since hospital discharge (including recovery, readmission, or death) are gathered from parents and may be supplemented by review of medical and administrative records. At the follow-up assessment, questions about knowledge, attitudes, and practices (KAP) toward influenza illness, including the extent of worry about influenza [30], are asked for all parents; KAP toward influenza vaccines are asked among parents of infants aged >6 months, including previously validated measures of intention to vaccinate [31–33] and the perceived effectiveness and safety of influenza vaccines [31, 33].

Community non-ill infants study design

For each study year, concurrent with hospital enrollment, approximately 150–300 non-ill infants aged <1 year per site are enrolled during routine preventive care encounters with weekly enrollment goals stratified by ages 0–5 months and 6–11 months in order to reflect the mixture of ages observed among hospitalized infants. As described in Table 2, all study sites enrolled non-ill infants during routine immunization visits; Jordan also enrolled infants during routine growth monitoring and other well-baby check-up visits; Nicaragua enrolled additional non-ill infants from an existing prospective cohort study [34, 35]. Screening questions confirm that the infant has not been acutely ill within the past 7 days with nasal congestion or discharge, new or worsening cough, shortness of breath, difficulty breathing, wheezing, fever, chills, or diarrhea. After informed consent is obtained from parents, trained staff collect combined nasal and OP swabs and complete a structured interview covering the same infant, household, maternal characteristics, and KAP assessed for hospitalized infants (Table 3). No blood is collected from non-ill infants. Starting in the second year of the study, a follow-up contact will be made to parents of non-ill infants (4–10 days after enrollment) to assess whether the infant developed illness symptoms after the initial interview.

Laboratory methods

Respiratory specimens are tested for influenza and RSV by reference laboratories at each study site using validated singleplex real time reverse transcriptase polymerase chain reaction (rRT-PCR) assays, with protocols, primers, probes, and reagents supplied by US CDC.Footnote 10 Study laboratories completed WHO influenza proficiency panels and quality assurance testing (of every 5th influenza positive and every 20th influenza negative) administered by Marshfield Clinic Laboratory (Marshfield, Wisconsin). RSV proficiency panels were administered and reviewed by US CDC.

Also at each study site’s research laboratory, we will complete multiple virus testing on respiratory specimens from both hospitalized and non-ill infants using multiplex real-time PCR for detection of pathogen genes by TaqMan® technology (Fast-track Diagnostics [FTD], Silema, Malta); specifically, we are utilizing FTD Respiratory Pathogens 21, which is a five tube multiplex for detection of influenza A virus, influenza A(H1N1) swine virus, influenza B virus, rhinoviruses, coronaviruses (NL63, 229E, OC43, HKU1), parainfluenza viruses 1-4, human metapneumovirus, bocavirus, respiratory syncytial virus, adenoviruses, enteroviruses, parechoviruses, Mycoplasma pneumoniae, and an internal control. Prior to beginning multiplex testing, each site laboratory completed Respiratory II EQA (External Quality Assessment) Programme by QCMD (Glasgow, Scotland; RESPII16 Catalogue Number QAV164189), which includes molecular detection and determination of human metapneumovirus, respiratory adenoviruses, rhinoviruses, coronaviruses and parainfluenza viruses. Although we will compare influenza and RSV results from both singleplex and multiplex assays, the singleplex rRT-PCR results will be the “gold standard” or primary assay.

A random subset of all influenza viruses may undergo virus isolation, depending on the availability of virus characterization information from local surveillance platforms. Also, some influenza viruses/specimens may have molecular characterization with genetic sequencing and other assays to detect genetic markers and antiviral resistance. A sampling of specimens positive for RSV will be selected and characterized by target genetic sequencing. Remaining aliquots of all study respiratory specimens will be sent to a CDC-designated facility for banking and storage; no specimens will contain personal identifiers. These specimens may also be utilized to investigate biomarkers of disease severity or novel viruses or pathogens.

Hemagglutination inhibition (HI) assay will be completed following the US CDC protocol [36] at the Battelle laboratory (Aberdeen, Maryland), after completing a proficiency panel. Duplicate results will be used to calculate the geometric mean antibody titer (GMT) against seven influenza A and B virus strains that include local influenza vaccine strains and circulating strains for the study year (for possible infant exposure) and prior years (for possible maternal exposure). Serologic confirmation of influenza infection will be defined as ≥4 fold rise in GMT from acute to convalescent sera, with convalescent HI titer of ≥ 40 GMT.

RSV-specific serum IgG, IgM, and IgA will be detected using standard enzyme-linked immunosorbent assay (ELISA) procedures [37, 38]. Endpoint titers will be determined for serum antibodies to the F and G proteins of RSV.

Data management

Data collection and site-level management are conducted using REDCap (Research Electronic Data Capture), which is a browser-based metadata-driven software system (Vanderbilt University, Nashville, TN) [39]. The recruitment log (including randomization and prioritization by recruitment strata), screening survey, informed consent, enrollment interview, specimen management, laboratory tracking and results, daily hospital abstraction, and follow-up assessment occur via REDCap forms on personal computers or mobile devices (tablets or smart phones).

Routine quality assurance monitoring is conducted at each study site and centrally by the data coordinator (Abt Associates). Missing or unclear information is corrected through follow-up contact with parents or medical record abstraction. Chained imputation methods or other methods will be used for imputing missing covariate data (i.e., not primary outcomes), if necessary.

Statistical considerations

Sample needs for the objectives to assess the percentage of influenza and/or RSV positives (by molecular or serologic methods) are driven by the precision with which we wish to measure the lowest expected percentage of infection for a particular category of infant hospitalizations at a specific study site. For optimum precision, we intend that the 95% confidence interval (CI) is within the range of plus or minus 50% of the observed percentage positive. For example, in order to measure a percentage of rRT-PCR confirmed influenza of 5% (with a 95% CI from 2.5–7.5%) within a particular patient strata would require a sample size of 328 (and 16 positives) [40]. Thus, we expect to make site-specific estimates for large infant categories (e.g., respiratory vs. non-respiratory illnesses). However, it will be necessary to combine results across study sites in order to estimate positivity for combinations of categories (e.g., non-respiratory illnesses among infants aged 0–5 months in ICU).

Sample goals for the non-ill infants were originally made with the expectation of observing at least 2% of infants positive for influenza virus and RSV infections [27, 41–43]. We recognized that 200 non-ill infants would be needed to observe a 2% rate of virus positivity with minimum precision (with a 95% CI = +/- 1.9%). Estimation of the attributable fraction of disease associated with influenza or RSV will require statistical adjustment and/or matching of ill and non-ill infants by season, weeks within season, age, and characteristics that may differ between infants recruited from preventive care settings and the source population [27]. The proposed 150–300 non-ill enrollees per year may not be adequate to capture this variation adequately.

Ethical approval & ethical considerations

The study protocol and procedures have been reviewed and approved by Institutional Review Boards (IRBs) at each study site and by Abt Associates (the coordinating institution on which US CDC relies). Written informed consent is completed in the parent or guardian’s language (Spanish, Arabic, Albanian, or Visayan [Boholano dialect]).Footnote 11 Provision of a small gift or travel reimbursement vary by study site depending on local IRB guidelines. Whenever possible, blood collection is coordinated with clinically necessary blood draws. Given the research nature of the laboratory methods and time delays in batch testing, rRT-PCR findings are not available to inform clinical decisions.

Discussion

We described the recruitment, data gathering, laboratory, and follow-up procedures for this prospective study of approximately 3,000 hospitalized infants and 1,400 non-ill infants aged <1 year as part of the US CDC-funded IRIS network. Estimating the hospital burden of severe infant disease associated with influenza and RSV and expanding our understanding of the clinical presentation and antibody response associated with these diseases are especially timely and relevant aims given existing and new immunoprophylaxis options for RSV [44, 45], the existing availability of influenza vaccines, and the development of new types of influenza and RSV vaccines and vaccine strategies to provide direct or indirect protection for infants [46–49].

Study strengths

By focusing on all acute infant illnesses, our prospective study promises to quantify the proportion of severe influenza and RSV disease that is missed by traditional surveillance approaches, case definitions, and molecular diagnostics. It is unique in its systematic sampling of acutely ill infants that does not rely on clinical judgment, its collection of in-depth clinical and socio-demographic information, and its use of both molecular and serologic diagnostics. Our Network extends research beyond the USA and other high-income countries and also includes countries with prolonged or multiple epidemic periods (Nicaragua and the Philippines), which are also under-represented in influenza and RSV research. Other strengths include the multi-country composition of our network, which will allow us to examine the consistency of trends across settings with different medical systems and infant populations. In addition, given the multi-agent source of many infant diseases [27, 50–55] and the absence of community controls in many hospital studies [27], the enrollment of non-ill infants from the same community is an important study feature which will allow us to examine the frequency of influenza virus and RSV infections among non-ill infants and, depending on sample size and virus attack rate, possibly estimate the virus-specific attributable fraction of influenza and RSV to severe disease.

Limitations

Our study has at least six limitations. First, the number and types of influenza virus infections we identify will depend on the circulating viruses in each country during the years of our study and may under- or over-estimate the expected frequency of disease in future years. Indeed, reviews of influenza studies in Africa and India observed 5- to 20-fold differences in incidence estimates across studies [56, 57]. Similarly, due to the timing and funding cycle of the study, we may not enroll infants during complete influenza and RSV seasons in all countries for all study years.

Second, different hospitals and medical systems likely have different clinical thresholds for hospitalization [6, 27, 58, 59], which will make comparisons across settings difficult; yet, this may be less of a concern when we examine the subset of infants requiring intensive care, who would likely be admitted in most settings. Third, the IRIS network does not include a low income country. Therefore, our ability to generalize from the middle-income countries in the IRIS network to low-income countries, where the burden of infant disease is high and access to hospital care is poor, is limited.

Fourth, although we expect our study methods will increase the identification of influenza and RSV infections, we recognize that our observations will continue to underestimate the true burden of these diseases; for example, serologic diagnostic assays will fail to identify some infections due to known limits in the measurement sensitivity of HI [60–62] and the inability of immune challenged infants to mount a robust antibody response [63–65]. Fifth, as an observational study that relies on existing local clinical practices, we will not have standardized measures of vital signs or clinical laboratory values using common procedures and calibrated instruments, which will limit the validity, reliability, and comparability of these indicators. Sixth, our focus on viral etiologies excludes description of other infectious and non-infectious etiologies of acute illness.

Conclusions

The IRIS network promises to expand our knowledge of the frequency, clinical features, and antibody responses associated with serious influenza and RSV disease among infants aged <1 year. Although variations in findings across study sites and years is a common challenge [66], we hope to identify at least two consistent trends regarding influenza among infants: (i) the proportion of non-respiratory hospital admissions among infants that are associated with influenza infections; (ii) the proportion of influenza infections that are missed by molecular diagnostics but can be identified by serologic diagnostics. For RSV, we aim to: (iii) identify risk factors associated with severe RSV disease across countries; (iv) evaluate the RSV-specific antibody response among infants hospitalized with severe RSV infections. These insights could aid countries in making more precise estimates of disease burden and inform decisions about the potential value of existing and new vaccines and other prevention and treatment strategies.