The Clinical and Genomic Epidemiology of Rhinovirus in Homeless Shelters—King County, Washington 

Abstract

In the United States, almost 600 000 people experienced homelessness nightly in 2020 [1] with approximately 11 751 people experiencing homelessness (PEH) in King County, Washington alone [2]. The ongoing coronavirus disease 2019 (COVID-19) pandemic has highlighted the health risks posed by respiratory viral infections in PEH. PEH have a disproportionate burden of chronic disease, exacerbated by mental illnesses, substance use [3], and social inequities [4], leading to an increased risk of premature mortality [5]. PEH who stay in shelters are at increased risk of infection due to difficulties with limited space for social distancing, isolation of sick individuals, contact tracing, adequate ventilation, and sanitation [6, 7]. Despite the public health challenges posed by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in homeless shelters, respiratory virus epidemiology, including rhinovirus (RV), in these settings remains poorly understood.

RV cocirculates with other respiratory viruses contributing to the global burden of respiratory diseases [8]. Prepandemic surveillance in the United States demonstrated year-round RV circulation with seasonal peaks in the spring and fall [9]. Although referred to as a cause of the common cold [10], RV infections in both children and adults can result in lower respiratory tract infections and exacerbations of underlying conditions, including asthma or chronic obstructive pulmonary disease (COPD) [11]. RV includes 3 major viral species (RV-A, RV-B, RV-C) with 160 known types, hindering efforts to develop viable vaccine candidates [12]. Thus, the focus remains on nonpharmaceutical measures to reduce RV burden. During the COVID-19 pandemic, RV continued circulating despite mitigation measures that have interrupted circulation of influenza, respiratory syncytial virus (RSV), and many other viruses [13]. In this study, we describe the epidemiology of RV infections in homeless shelters in King County, Washington before and during the first year of the COVID-19 pandemic. We use genomic sequencing to characterize the molecular RV diversity to understand the nuanced complexities of RV epidemiology in shelter sites.

METHODS

Study Design, Setting, and Population

We retrospectively analyzed cross-sectional data from 2 studies: (1) a randomized control trial of influenza testing and treatment (NCT04141917) occurring October 2019–31 March 2020 and October 2020–31 March 2021, and (2) a SARS-CoV-2 surveillance study from 1 April 2020 onward. Details of the methods of these studies have been previously described [14, 15]. Participants were enrolled at staffed kiosks from 23 homeless shelter sites within King County, Washington from October 2019 to May 2021 and data from the enrollment questionnaire and respiratory samples were used for this study. Briefly, between October 2019 and 31 March 2020, eligible participants were shelter residents aged ≥3 months with the following symptoms in the last 7 days: new or worsening cough or at least 2 symptoms including subjective fever, headache, sore throat, runny nose or congestion, shortness of breath, and muscle or body aches; for participants aged <18 years, diarrhea, rash, and ear pain or discharge were also included. Once a month, asymptomatic participants were permitted to enroll. With the community spread of SARS-CoV-2, participant enrollment eligibility expanded to include shelter residents and staff regardless of symptoms from 1 April 2020 onward for all studies. As part of Public Health—Seattle and King County contact tracing efforts, 1-day large-scale (surge) testing events were implemented within shelter sites with a SARS-CoV-2–positive case.

Consent was obtained from participants aged ≥18 years or from a guardian for those aged <18 years; assent was obtained from participants aged 13–17 years. At enrollment, participants submitted questionnaires and a respiratory sample for respiratory virus testing. Study enrollment was limited to weekly participation except in cases where new or worsening symptoms developed. Multiple enrollments from the same participant were linked by participant name and birthdate. Encounters refer to each time the participant enrolled in the study. This article was prepared using deidentified study data. The study was approved by the University of Washington Institutional Review Board (study 00007800).

Questionnaire, Variables, and Shelter Site Data

After study consent, the study team administered an enrollment questionnaire via electronic tablet. Questionnaire data, including shelter site, birthdate, sex, race, ethnicity, symptoms, pregnancy status, underlying medical conditions, and current tobacco use (including e-cigarettes), were stored through Research Electronic Data Capture (REDCap). Underlying medical conditions collected by self-report included neurological disease, cardiovascular disease, asthma, bronchitis, COPD, hepatic disease, diabetes mellitus, immunosuppression, cancer, or another condition that was not listed. New or worsening illness course symptoms over the last 7 days were collected in the questionnaire: runny nose or congestion, cough, sore throat, fatigue, myalgias, headaches, subjective fevers, shortness of breath, sweats, nausea or vomiting, chills, diarrhea, rash, ear pain or discharge, and loss of taste and smell (added after 1 April 2020). Encounters where no new or worsening symptoms were reported were classified as asymptomatic. We defined influenza-like illness (ILI) as reported fever and either cough or sore throat and COVID-19–like illness (CLI) was defined as reported fever and either cough or shortness of breath. We obtained resident demographics that the shelter served from shelter management staff.

Specimen Collection and Respiratory Virus Testing

Respiratory samples were collected at enrollment. From the start of the study to 22 July 2020, and then from 1 November 2020 through study end, samples were obtained via midturbinate sterile nylon flocked swabs. Anterior nares swabs were used from 22 July 2020 through 1 November 2020 due to supply chain limitations. Specimens were initially collected by study staff, but staff supervised self-collected swabs were used from 6 March 2020, with the community spread of SARS-CoV-2 necessitating heightened safety measures for staff. Respiratory viruses were detected using a custom arrayed reverse transcription polymerase chain reaction (RT-PCR) platform (Thermo Fisher Open Array) including: influenza virus (A, B, and C), respiratory syncytial virus (A and B), human parainfluenza (1–4), human coronaviruses (HCoV-OC43, HCoV-NL63, HCoV-HKU1, HCoV-229E), RV, enterovirus, human bocavirus (excluded after 29 May 2020), human parechovirus (excluded after 23 November 2020), human metapneumovirus, and adenovirus. Due to the potential for cross-reactivity between RV and enterovirus, we used a custom review process to differentiate between these viruses (Supplementary Methods). Specimens from 1 January 2020 onward were tested for SARS-CoV-2. Details of SARS-CoV-2 testing have previously been published [15]. RV codetection was defined as RV detection with ≥1 other virus. For virus-positive samples, a relative cycle threshold (Ct) was calculated.

Genomic Sequencing and Analysis

RV whole-genome sequencing was attempted on RV-positive samples with Ct values <17 and a subset with Ct values ≥17. RNA was extracted using the Roche MagnaPure 96 DNA and viral nucleic acid small volume kit, Viral NA Universal SV 4.0 protocol (200μ input, 50μ elution). Shotgun metagenomic sequencing libraries were prepared as previously described [16, 17]. Raw reads were processed using a custom published pipeline [18]. Additional information is in Supplementary Methods.

Computational Analysis

We analyzed demographic and symptom data descriptively. We used SAS software version 9.4 for general data analysis. NextStrain software was used to process consensus genomes and for the assembly and visualization of phylogenetic trees [19]. Bootstrap values were calculated using IQ-TREE, version 1.6.12 [20]. In addition to the consensus genomes generated for this study (GenBank accession numbers: ON311150–ON311280; Supplementary Table 1), we downloaded and included in our analyses full-length RV genomes available from GenBank.

RESULTS

Between October 2019 and May 2021, there were 14 464 encounters (Figure 1) from 3281 unique participants (median age, 37 years; range, 0.3–85 years; 86% adults; 60% male; 40% white). Overall, 46% of participants reported smoking (of whom 16% reported e-cigarette use), 31% reported ≥1 underlying medical condition, and 17% were shelter staff. Among 14 421 encounters where the encounter date was known, 12 731 (88%) encounters occurred after 1 April 2020. There was a mean of 721 monthly encounters over the study period with a mean of 909 monthly encounters after 1 April 2020 (Supplementary Table 2). A total of 12% and 90% of encounters before and after 1 April 2020, respectively, involved participants who were asymptomatic at enrollment. There were 12 895 (89%) encounters with samples where no respiratory virus was detected with 83% involving asymptomatic encounters. Among all symptomatic encounters before and after 1 April 2020, 27% and 13%, respectively, of samples collected had ≥1 respiratory virus detected, of which 43% and 75%, respectively, were RV positive. Among all asymptomatic encounter before and after 1 April 2020, 16% and 9%, respectively, had ≥1 respiratory virus detected; of which 56% and 78%, respectively, were RV positive.

Figure 1.

Study flow diagram.

Open in new tabDownload slide

A mean of 53 RV-positive samples were collected monthly over the entire study period with RV-positive samples present every month from October 2019 to May 2021 (Figure 2). The percentage of RV-positive samples before and after April 2020 was 11% and 7% (a higher percentage than other viruses detected during this time), respectively. There was an increase in the proportion of RV-positive samples obtained from virus-positive asymptomatic participants (from 56% to 78% before and after April 2020, respectively) associated with enrollment symptom criteria expansion. RV was the most common respiratory virus throughout the study (Supplementary Table 3) with 66% involving adult participants and 10% shelter staff. RV was detected in 1066 samples (7.4% of all samples) from 682 unique participants (median age, 30 years; range, 0.3–85 years; 58% male; 42% white; Table 1) representing 68% of all virus-positive samples. RV was the only virus detected in 986 samples from 647 participants (median age, 29 years; range, 0.3–85 years; 58% male; 41% white).

Figure 2.

Frequency of rhinovirus-positive samples by participant encounter symptom status over the study period. n = 3 rhinovirus-positive samples with missing encounter dates were excluded from this figure. A symptomatic encounter was defined as a study encounter in which the participant reported any new or worsening symptom on the enrollment questionnaire and was not limited to symptoms required for enrollment; an asymptomatic encounter was defined as a study encounter in which the participant did not report any new or worsening symptoms on the enrollment questionnaire.

Open in new tabDownload slide

Participants in shelters housing families (adults and children) and young adults (18–25 years) had the highest prevalence of RV detection relative to other shelters, constituting 12% and 8% of all encounters from these sites, respectively (Table 2). Participants aged <5 years had the greatest proportion of RV-positive samples (26%) while participants aged ≥65 years had the lowest proportion compared to the other age groups (4%; Table 3). Viral codetection with RV occurred among 80 (8%) RV-positive samples (49% were adult encounters), with adenovirus being the most common codetected virus (36% of samples with RV codetection; Supplementary Table 4).

Among the 647 unique participants with only RV detected, 69% had asymptomatic encounters compared to 56% of the 66 unique participants with RV codetection (Table 4). Runny nose (79%), cough (61%), and sore throat (42%) were the most common symptoms reported by unique symptomatic adult participants with RV only while runny nose (65%), cough (58%), sore throat (26%), and nausea/vomiting (26%) were the most common symptoms in pediatric participants (Supplementary Table 5). Of note, 3 participants (2 adults and 1 child) with RV only reported new loss of sense of taste or smell, all of whom were tested for SARS-CoV-2 and did not have a positive or inconclusive SARS-CoV-2 test result. The proportion of unique symptomatic participants with RV infection reporting ILI and CLI symptoms was higher in those with RV codetection than with RV only (ILI, 24% vs 17%; CLI, 21% vs 16%, respectively). Among all encounters where ILI was reported, 9% had RV infection only, and among all encounters where CLI was reported, 9% had RV infection only. Among all symptomatic encounters, 9% of adults had RV detected while 26% of children had RV detected.

We generated full genome sequences for 131 of 176 RV-positive samples, including 24 with Ct value ≥17 (1 genome with approximately 23% missing data, all others with <10% missing data). Sequenced samples were collected from every month of the study period except for May–June 2020 and were from 10 different shelters. A total of 33 different RV types were represented among the sequenced samples: 14 RV-A types, 4 RV-B types, and 15 RV-C types. RV-A23 was most common (31 out of 131 sequenced samples) while 12 types were represented by only 1 sequence. Sequenced samples were collected across 18 months from October 2019 to May 2021, but no individual type was observed for >4 months (Supplementary Table 6). Of the 9 types observed before 1 April 2020, only 1 was also observed after this date when community-wide mitigation efforts were implemented.

Of the 33 total observed RV types, 14 originated from more than 1 shelter (Supplementary Table 7). RV-A23, RV-A34, and RV-B27 were all observed in 5 different shelters. Shelter D, a family shelter and the source of the most sequenced samples (n = 38), had the highest number of different RV types (n = 20) among its sequenced samples (Supplementary Table 6) and the highest number of types observed in a single shelter in 1 month (4 in January 2021). In addition to having the highest overall number of RV cases and sequenced cases, family and young adult shelters had cases due to more RV types than other adult shelters (range, 8–20 vs 1–4).

There were 27 instances where >1 sample of the same type was collected from the same shelter. Among these, there were 10 pairs of identical sequences, 4 sets of 3 identical sequences, and 1 set each of 4, 5, and 7 identical sequences, so that a total of 48 genomes were identical to at least 1 other genome from the same shelter. We constructed RV-A, RV-B, and RV-C phylogenetic trees, which included sequenced study samples and 947 RV-A, 201 RV-B, and 348 RV-C genomes from GenBank. Within these trees, 17 of 27 sets of genomes of the same type and shelter of origin clustered together exclusive of all other shelters and all GenBank genomes with good bootstrap support (≥89%; Figure 3 and Supplementary Figures 1 and 2). Figure 3A shows several examples in which this was not the case as RV-A23 samples from shelters C, D, and H formed more than 1 distinct phylogenetic grouping within this tree. While the 2 clusters for shelters C and D represented samples collected at different times, there was chronologic overlap in sample collection dates for the 2 largest shelter H clusters.

Figure 3.

Maximum likelihood phylogenetic tree of select rhinovirus types: (A) rhinovirus-A23, (B) rhinovirus-A34, and (C) rhinovirus-B27. Nodes are colored by the shelter of origin; GenBank samples are gray.

Open in new tabDownload slide

The relationship among sequenced genomes of the same type from different shelters varied across types. Fifteen types were observed in more than 1 shelter. For 6 of these 15 types, all shelter samples formed a monophyletic group exclusive of all GenBank genomes of that type, while for 5 of these 15 types, the minimum genetic distance between sequenced samples from 2 different shelters was <5 single nucleotide changes. This includes 2 pairs of identical sequences for which each sequenced sample came from a different shelter.

DISCUSSION

RV was the most common respiratory virus detected before and during the COVID-19 pandemic among individuals in homeless shelters in a major metropolitan region. There were RV-positive samples detected in every month during the study period. RV-positive samples were most common in younger age groups and among samples collected from shelters housing families (adults and children). Although RV was prevalent throughout the study period, the number of viral codetections was relatively low. Sequenced RV samples included >30 different RV-A, RV-B, and RV-C types, the relative frequencies of which varied significantly over the study period. Our findings show that despite the implementation of community-wide mitigation efforts, including the Washington State Stay-At-Home Ordinance [21], RV persisted in homeless shelters throughout the study period, a trend similarly found in studies during the COVID-19 pandemic period.

RV was a substantial contributor to the respiratory viral infections in individuals of all ages in homeless shelters in this study, a finding reported by others in congregate settings. In a respiratory pathogen study in homeless shelters in France prior to the COVID-19 pandemic, RV was similarly found to be the most detected respiratory virus [22]. Nursing homes are another congregate setting where RV infections are common. In one study of symptomatic individuals, RV was the most common virus in nursing home staff, more common than RV in residents [23]. Another nursing home surveillance study, from December 1989 to March 1990, found RV to be a common respiratory viral infection second only to RSV infections in residents with acute respiratory illness symptoms [24]. Direct comparison of RV frequency to these studies may be limited as study participants were mostly screened for the presence of symptoms and asymptomatic sample collection was limited. Our study adds to this congregate-setting literature by showing that symptomatic disease is only a subset of RV infections and that RV asymptomatic encounters in homeless shelters was common. What role individuals play in RV transmission in homeless shelters when asymptomatic is not known. Furthermore, we found that ILI and CLI syndromic surveillance definitions are insufficient to capture the full breadth of symptomatic RV encounters and more sensitive definitions are needed for assessment of RV burden. In congregate settings, transmission prevention between individuals may be difficult, placing those with comorbid factors at increased risk of clinical complications [25, 26]. Longitudinal studies in homeless shelters with clinical outcomes are needed to better understand the scope of RV-associated burden in these settings.

The combined effects of COVID-19 pandemic mitigation policies, including the local issuance of the Washington State Stay-At-Home ordinance on 23 March 2020 [21], on respiratory virus circulation continues to be an important area of study. With continuous enrollment throughout the study period, we found that RV detection persisted in the homeless shelter setting as the COVID-19 pandemic progressed. Similar findings were found in a French shelter study early in the COVID-19 pandemic, where only SARS-CoV-2 and RV were found over the study period [27]. In a California respiratory virus sentinel surveillance system study from May 2020 to June 2021, RV/enterovirus activity returned to near normal levels in the fall of 2020 after initial decreases spanning the spring and summer of 2020 [28]. A national US surveillance study showed an overall decrease in number of specimens testing positive for non–SARS-CoV-2 respiratory viruses early in the pandemic [13]. Despite an initial decrease, RV/enterovirus increased back to levels seen before the pandemic from May 2020 onward. How RV case numbers rapidly returned back to prepandemic levels and persisted despite broad nonpharmaceutical interventions is likely multifactorial. Some explanations include prolonged RV shedding [29], ease of reinfection given type diversity [30], viral interference [28], transmission from contacts and fomites [31], lower efficacy of face masks in respiratory spread compared to other respiratory viral infections [32], and environmental resistance as a nonenveloped virus [29]. In shelters, suboptimal ventilation may also contribute to RV persistence. Studies in closed environments have demonstrated effective aerosol transmission of RV [33, 34]. These findings show the importance of additional virus-specific studies to identify the factors that affect their unique epidemiology.

Genetic sequencing in a subset of RV-positive samples illustrated the diversity of RV infections in these shelter sites, including 20 types observed in 1 shelter site alone. RV types identified prior to the implementation of community-wide mitigation policies were largely not observed from 1 April 2020 onward. Across the study period, individual RV types were observed for limited periods of time (<4 months) before being replaced by other types. Despite the RV type diversity seen, there were also multiple examples where samples of the same type were collected from participants in the same shelter. These samples frequently formed phylogenetic clusters exclusive of other shelter sequences and over a third of all sequenced shelter samples were identical to at least 1 other sample from the same shelter, observations which may be indicative of intrashelter spread. Two instances of identical sequence pairs collected from 2 different shelters also raises the possibility of RV spread between shelters. However, our ability to assess this is limited, given the lack of RV samples collected in the surrounding community during the study period. Finally, our data suggest that multiple introductions of the same viral type into one shelter in a short time period is possible (RV-A23 in shelter H). Overall, the genetic diversity of RV in our study sites highlights the importance of including RV sequencing analysis in studies of RV epidemiology in this population.

Our study was subject to several limitations. First, there may have been an underestimation of RV-positive samples as non–pan-RV primers were used in the RT-PCR assay. Second, selection bias may have occurred through participant self-recruitment. Third, participants were not followed longitudinally thus limiting our ability in differentiating asymptomatic infection from presymptomatic participants or from those with persistent shedding after symptomatic infection. Fourth, our study did not collect shelter site nonpharmaceutical interventions that were implemented over the course of the COVID-19 pandemic, limiting inference on how they may have affected respiratory viral transmission. Fifth, we used anterior nares swabs between July 2020 and November 2020, which may have reduced sensitivity for respiratory viral detection over this time period. Sixth, despite utilizing an algorithm to differentiate RV and enterovirus-positive samples, there may have been misclassification in samples not sequenced. Seventh, human bocavirus and human parechovirus were not tested for towards the end of our study and samples over that period may have missed those viruses. Finally, we were able to perform genomic sequencing only on a subset of RV-positive samples and so it is likely that the diversity of RV types is not completely described.

CONCLUSIONS

RV is an important viral pathogen in homeless shelters affecting individuals of all ages. Similar to observations nationally, RV cases and diversity persisted in our study despite COVID-19 community-wide mitigation efforts. RV genomic analysis suggested that both intrashelter spread and new introductions into shelters were common and impacted persons of all ages. Respiratory viral epidemiology, including RV, present unique public health challenges in congregate settings. Future congregate-setting–based studies of RV surveillance and transmission as pandemic interventions change can build upon the findings in our study.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We thank all the shelter staff, managers, and participants for their collaboration and help with our research study. We acknowledge Amy C. Link, Ashley Kim, Jessica O’Hanlon, Jessica Heimonen, Naomi Wilcox, and Jennifer Logue along with all the research assistants who contributed to the success of this study. We thank Dr Michael Boeckh and Mr Luis Gamboa for their scientific contributions to this study.

Disclaimer. The views expressed are those of the authors and do not necessarily reflect the official policy of the Centers for Disease Control and Prevention.

Financial support. This work was supported by Gates Ventures; the Centers for Disease Control and Prevention (contract number 75D30120C09322 AM002 to H. Y. C.); and the National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases (grant number T32 AI007044 to E. J. C.). Computational analyses were supported by Fred Hutch Scientific Computing (NIH Office of Research Infrastructure Programs grant number S10OD028685) and University of Washington Laboratory Medicine Informatics.

Supplement sponsorship. This article appears as part of the supplement “Homelessness and Infectious Diseases: Understanding the Gaps and Defining a Public Health Approach,” sponsored by the Centers for Disease Control and Prevention.

References

Author notes