Home collection of nasal swabs for detection of influenza in the Household Influenza Vaccine Evaluation Study

Abstract Background Community‐based studies of influenza and other respiratory viruses (eg, SARS‐CoV‐2) require laboratory confirmation of infection. During the current COVID‐19 pandemic, social distancing guidelines require alternative data collection in order to protect both research staff and participants. Home‐collected respiratory specimens are less resource‐intensive, can be collected earlier after symptom onset, and provide a low‐contact means of data collection. A prospective, multi‐year, community‐based cohort study is an ideal setting to examine the utility of home‐collected specimens for identification of influenza. Methods We describe the feasibility and reliability of home‐collected specimens for the detection of influenza. We collected data and specimens between October 2014 and June 2017 from the Household Influenza Vaccine Evaluation (HIVE) Study. Cohort participants were asked to collect a nasal swab at home upon onset of acute respiratory illness. Research staff also collected nose and throat swab specimens in the study clinic within 7 days of onset. We estimated agreement using Cohen's kappa and calculated sensitivity and specificity of home‐collected compared to staff‐collected specimens. Results We tested 336 paired staff‐ and home‐collected respiratory specimens for influenza by RT‐PCR; 150 staff‐collected specimens were positive for influenza A/H3N2, 23 for influenza A/H1N1, 14 for influenza B/Victoria, and 31 for influenza B/Yamagata. We found moderate agreement between collection methods for influenza A/H3N2 (0.70) and B/Yamagata (0.69) and high agreement for influenza A/H1N1 (0.87) and B/Victoria (0.86). Sensitivity ranged from 78% to 86% for all influenza types and subtypes. Specificity was high for influenza A/H1N1 and both influenza B lineages with a range from 96% to 100%, and slightly lower for A/H3N2 infections (88%). Conclusions Collection of nasal swab specimens at home is both feasible and reliable for identification of influenza virus infections.


| INTRODUC TI ON
Influenza is a respiratory virus that causes substantial annual morbidity and mortality, including an estimated 200 000 hospitalizations and 30 000 deaths in the United States each year. 1 Vaccines are available for prevention of influenza virus infections, but recent estimates have shown only moderate vaccine effectiveness (VE). 2,3 Annual variation is common, in terms of the frequency and severity of infection as well as in VE. As a result, further studies of influenza VE and transmission are needed with the goal of improving control.
Prospective, longitudinal community-based studies have a broad range of applications in respiratory virus epidemiology. These studies will be essential to better understand the extent of the pandemic caused by the novel coronavirus SARS-CoV-2 and the full range of COVID-19 illness. These studies also present unique opportunities to explore more in-depth questions about immune correlates of influenza vaccine failure as well as susceptibility to and transmission of infection. Nevertheless, they are much more resource-intensive than comparably sized studies using case-control designs. 4 Ensuring adequate and timely specimen collection across a large cohort is particularly important as the circulation of respiratory viruses varies greatly on both a seasonal and annual basis. The minimum detectable effect size for preventive interventions in these studies is particularly sensitive to variations in the infection risk. Sensitive methods of pathogen detection (eg, RT-PCR) have improved identification of cases, but specimen collection methods that are timely and reduce the burden on study participants are needed to minimize the likelihood of incorrectly determining infection status. 5 There have been several feasibility and validation studies which have suggested that self-or parent-collected nasal swabs are both acceptable and result in quality specimens for identification of respiratory viruses. [5][6][7][8][9][10][11] There are few studies, however, involving community-based participants collecting specimens in their own homes, outside of a study clinic setting. 12 We sought to describe the feasibility of home-collected respiratory specimens collected at home during an acute respiratory infection (ARI) and examine the validity of the specimens compared to those collected by research staff.

| ME THODS
We collected respiratory specimens at two time points during ARI to

| ARI surveillance
Active surveillance for identification of ARI was conducted year round, beginning in October 2014. Participating households were instructed to contact the study team at the onset of new ARI and were additionally contacted each week by email or telephone.
Report of an illness meeting the study case definition triggered collection of an upper respiratory specimens at home on the first day of symptoms (home-collected) and then in the study clinic within 7 days of symptom onset (staff-collected).

| Home-collected respiratory specimens
Participants (or their parents) were asked to collect a nasal swab specimen at home on the first day of their illness (home-collected specimens). Adults were trained in-person to collect specimens prior to illness season at enrollment visits. In addition, each household was given an instruction card and a link to an online video with detailed instructions on how to collect nasal swab specimens. Home-collected specimens were stored at room temperature in commercially prepared viral transport media and submitted to research staff during their scheduled illness visit in the study clinic.

| Staff-collected respiratory specimens
Research staff scheduled a specimen collection visit at the study clinic within 7 days of illness onset. At these visits, oropharyngeal and midturbinate swabs (mid-turbinate only in children <3 years of age) were collected by research staff and combined in commercially prepared viral transport media (staff-collected specimens). Specimens were kept at room temperature until they were transported to laboratory.

| Influenza testing
Specimens were tested by RT-PCR for laboratory confirmation of influenza, using primers and probes from the Centers for Disease Control and Prevention. Influenza subtype was determined for influenza A-positive specimens, and lineage was determined for influenza B-positive specimens. Specimens were also tested for human K E Y W O R D S acute respiratory illness, community-based, influenza, self-collected nasal swabs ribonuclease P (RNP), using a cycle threshold (Ct) cutoff of ≤40 to determine specimen quality. 13,14 Specimens without detection of RNAseP were excluded from further analyses.
All staff-collected specimens were tested for influenza by RT-PCR. We selected a subset of paired home-collected specimens for influenza testing if staff-collected specimen was RT-PCR-confirmed influenza-positive, if staff-collected specimen test results were inconclusive for subtype or if the onset of symptoms was within 7 days of an influenza case in a household contact. In all cases, the specimen collected by staff was within 7 days of onset.

| Statistical analysis
We first described the proportion of ARI with home collection of specimens by season, and participant and household characteristics. We examined the timing of home collection and clinic collection in relation to the onset of symptoms. Finally, we examined the reliability of home-and staff-collected specimens. Dichotomous outcomes, including detection of influenza, were compared by calculating Cohen's kappa. We interpreted the kappa statistic as sug- almost perfect. 15 We also examined the sensitivity and specificity of home-collected specimens using staff-collected specimens as the gold standard. Mean cycle threshold (Ct) values among concordant influenza-positive specimens were compared using the paired Wilcoxon test.  ARI events compared to the overall study population (Table 1).

| Study population
Approximately 70%-80% of ARI with staff collection of respiratory specimens in each year also included home collection of nasal swabs.
The subset of 336 paired home-and staff-collected specimens that were selected for influenza testing were representative of all ARI with specimen collection (Table 1).

| Timing of collection
Among the paired specimens tested for influenza, 313 (93%) were collected within 2 days of illness onset. The median duration from onset to home collection of respiratory specimens was 0 days (IQR 0-1) compared to 2 days (IQR 1-4) for staff-collected specimens ( Figure 1). Sixty-seven (20%) home-collected specimens were collected on the same day as staff-collected specimens. 109 (32%) home-collected specimens were collected 1 day prior to staffcollected specimens. The remaining 160 (48%) were collected ≥2 days prior to staff-collected specimens. Human RNaseP gene was detected at a Ct of less than 40 in all tested home-collected specimens and in all but one staff-collected specimens. The staffcollected specimen with RNaseP Ct >40 was excluded from further analysis.

| Influenza testing
We identified 150 cases of influenza A/H3N2 infection in staffcollected specimens. 123 of these infections were confirmed by testing home-collected specimens. An additional 22 cases of influenza A/H3N2 infection were identified by testing home-collected specimens (Table 2). Similarly, we identified 32 cases of influenza B/ Yamagata virus infection in staff-collected specimens. 25 of these were confirmed, and an additional 12 cases were identified by testing home-collected specimens. We also identified 23 cases of influenza A/H1N1 and 14 cases of B Victoria ( We also examined the validity and reliability of home-collected specimens by age-group. In these age-stratified comparisons, we combine all influenza types and subtypes due to small sample sizes. Agreement was generally consistent with the results for influenza A/H3N2, as that was the dominant subtype. We found lower agreement among adults and children aged 5-11 years and moderate agreement among younger children and adolescents (Table 3).
Sensitivity and specificity of home-collected specimens compared to staff-collected specimens were reasonably high for all age-groups.

| RT-PCR cycle threshold (Ct) values
Overall, the median Ct for staff-collected specimens was 24.4 (IQR

| D ISCUSS I ON
We found that unsupervised, home collection of respiratory specimens for identification of influenza infection was both feasible and reliable. A high proportion of participants completed the home collection and the vast majority completed collection within 2 days of symptom onset. Overall, the percent agreement was high for all influenza subtypes examined in this study. Agreement between home and staff collection was moderate for influenza A/H3N2 and B/Yamagata viruses and higher for A/H1N1 and B/Victoria viruses.
When stratified by age-group, agreement was generally consistent with the results for influenza A/H3N2, but was lower among adults.
Importantly, when specimens were collected within 1 day of each other, the agreement between collection methods was substantially higher. Additionally, assessment of RNAseP detection suggested that specimen quality was adequate for home-collected specimens.
Previous studies have demonstrated that nasal swabs collected by research participants under the supervision of research or F I G U R E 2 Ct value for home-collected and staff-collected respiratory specimens by influenza type, stratified by time (days) between home and staff collection clinical staff are feasible for the identification of respiratory virus infections. [6][7][8][9]12,16,17 In many of these studies, acceptability of self-or parent collection was preferred to investigator or clinician collection. 7,16 In the few studies of unsupervised collection, self-collected specimens were collected and returned in the timeframe recommended by the research team. 6,16 In addition, specimen quality in many studies has been similar regardless of collection method. 5,19 A Canadian study found that both viral loads and RNaseP were similar between self-collected and investigator-collected specimens. 8 A study in pregnant women found 100% of self-collected specimens detected RNaseP, 6 as we found in the current study. Other measures of specimen quality have also been similar between collection methods. 11,18,20 Importantly, many of these feasibility studies were pilot studies with relatively small sample sizes, and self-collection was often completed in the presence of the study investigators. 21 Our study confirms the feasibility and timeliness of unsupervised collection of respiratory specimens in a large, longitudinal cohort study. As we are currently experiencing with COVID-19 pandemic, unsupervised collection is essential to provide maximum protection to research staff and participants and to comply with social distancing guidelines from public health authorities.
High levels of agreement and high sensitivity and specificity of selfcollected specimens have also been demonstrated in a variety of settings. For example, self-collected nasal swabs have been compared to nasal wash with high levels of sensitivity (88%-95%). 9 A study of children < 5 years comparing supervised parental collection to pediatrician collected specimens found similarly sensitivity (89%) and high specificity (97%). 7 Additional studies have established that self-collected specimens are adequately sensitive and specific for detection of respiratory viruses. 8,17,19 Importantly, our study was not a random sample of paired specimens. Nevertheless, a random selection process would likely only change the prevalence of influenza as an outcome in these paired specimens. Therefore, we would not expect the sensitivity and specificity to differ with a random selection process. We observed similarly high validity of home-collected specimens in terms of sensitivity and specificity, compared to previous work.
The overall agreement was lower in our study compared to others 7,8,11 for influenza A/H3N2 viruses in part because we identified 29 additional influenza infections in participants who were influenza negative according to their staff-collected specimens.
Self-and staff-collected specimens in our study differed by both site and timing of collection in our study. But, recent work has suggested that there is very little difference in the sensitivity of nasal swabs compared to throat swabs 5,22 and that throat swabs added little in terms of respiratory virus detection when compared to nasal swabs alone. 23 The difference in agreement with previous studies may be explained by the timing of our collection methods. Many of the previous studies involved concurrent collection of self-and staffcollected specimens, whereas our specimens are collected on onset (home-collected) and again within 7 days of onset (staff-collected).
One study of parent-collected specimens found that time from symptom onset to collection was the only factor associated with respiratory virus positivity; collection method and subjective quality of parent-collected specimens were not associated with detection. 6 Likewise, we found that time between home and staff collection was an important factor in terms of both Ct value and agreement.

ACK N OWLED G EM ENTS
The authors wish to thank the HIVE Study participants for their participation in this research, as well as the HIVE Study staff for their hard work and dedication to the project.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.