Alterations in RSV Epidemiology in the United States following the COVID-19 Pandemic
To assess how RSV epidemiological trends changed in the US before and after the COVID-19 pandemic, we calculated 3-week rolling averages of diagnostic tests administered, RSV detections, and percent positivity from the National Respiratory and Enteric Virus Surveillance System (NREVSS) from July 2010 to April 2023 for all participating US laboratories (Fig. 1A). Diagnostic testing for RSV through the 2019–2020 season followed a regular seasonal pattern with the total number of tests administered peaking near 40,000 tests per three-week rolling average in later years. Detections and percent positivity rate followed similar seasonal trends, with most years peaking around 15 to 20% positivity during December or January. Testing increased as expected early in the 2020–2021 season, but detections and percent positivity remained near zero. Hereafter, the number of tests administered remained high year-round with additional seasonal testing during the winter months pushing peak testing to over 140,000 tests per three weeks in early 2023. This increased testing helped capture the broad off-season peak during the summer and fall of 2021, followed by an early and dramatic peak in detections the following season in November of 2022. Despite the altered seasonality and testing patterns, percent positivity remained in line with historic trends, peaking near 15 to 20%. Likewise, regional trends where the single-center study took place recapitulated the aberrant RSV behavior observed in national data (Supplementary Fig. 1).
We then leveraged hospitalization data from the RSV-Hospitalization Surveillance Network (RSV-NET)21 to determine whether increased viral detections coincided with increased hospital admission rates. Hospitalization data for patients with RSV were obtained from 12 states (California, Colorado, Connecticut, Georgia, Maryland, Michigan, Minnesota, New Mexico, New York, Oregon, Tennessee, and Utah) participating in the Emerging Infections Program22 or the Influenza Hospitalization Surveillance Program between July 2017 and April 2023.23 Irrespective of age, admission rates were below 1 case per 100,000 in the 2020–2021 season, reemphasizing the lack of RSV circulation at that time (Fig. 1B). Infants 0 to 4 years of age displayed a broadened and off-season peak in hospitalizations in 2021 followed by a heightened peak in October 2022, approximately 3–4 times higher than in prior years. These trends are less apparent in other age groups, which showed a more typical seasonality, particularly for older adults aged 65 and older. Similarly, while hospitalizations increased for other age groups in the 2022–2023 season, the magnitude of the increase was not proportional to what was seen for infants. Taken together, these data confirm alterations in RSV epidemiological patterns following the COVID-19 pandemic but suggest these trends were largely driven by infants as opposed to older adults.
Shifts in RSV Testing Platforms Resulted in an Incidental Increase in Outpatient Detections
National RSV-associated hospitalizations did not increase proportionately with cases in the 2022–2023 season, suggesting an increase in less severe case detections. We hypothesize that the less severe case detection trend will likewise be observed within our single-center study. To test this hypothesis, we performed a single-center retrospective cohort study of patients who tested positive for RSV at a Northwestern Medicine (NM) affiliated hospital or clinic between August 1st, 2009, and March 1st, 2023. clinical data collated through NM’s centralized repository system identified 8,508 unique patient encounters over this timeframe, including 3,416 adult and 5,092 pediatric patients (Supplementary Fig. 2). Most NM patient encounters were geographically concentrated in the greater Chicago area (Cook County and the surrounding suburbs) (Fig. 2A). Within Cook County, encounters were aggregated in the north and far north regions of the city with lower representation in the southwest and far southwest regions. To verify that RSV behavior in Chicago was representative of broader regional and national trends, we obtained 3-week rolling averages of diagnostic tests administered, RSV detections, and percent positivity from the Chicago Department of Public Health (CDPH) from September 29th, 2019, to January 22nd, 2023 (Fig. 2B). As expected, these data confirmed aberrant seasonality and testing patterns citywide following the COVID-19 pandemic with increased detections in the 2022–2023 season.
We assessed patient outcomes (i.e., outpatient, hospitalized, ICU admission, and death) and diagnostic testing platform information for the 3,416 adult patient encounters from our clinical dataset. Adult patient encounters increased gradually from the 2009–2010 season through the 2019–2020 season (Fig. 2C). These encounters were predominantly inpatients (> 75%) until the 2015–2016 season when encounters became more balanced at around 50% for both outpatient and inpatient encounters (Fig. 2D). After only a single observed encounter in the 2020–2021 season during the COVID-19 pandemic, cases rebounded to near-normal levels in 2021–2022 before doubling in the 2022–2023 season. These latter seasons had increased outpatient encounters, with over 75% outpatient proportion during the 2022–2023 surge. Notably, outcomes among inpatients varied only slightly during the study period, with a consistent 25–40% of inpatients requiring intensive care or dying each year (Fig. 2D, bottom panel).
These data suggest that most of the infected adult patients were detected in the outpatient setting. Historically, many adult RSV cases in the NM system have been detected using either a broad, multiplexed, PCR-based respiratory panel or a similar PCR-based assay for the detection of RSV and influenza viruses (Fig. 2E). After the start of the COVID-19 pandemic, however, platforms became predominated by multiplex tests for a range of viruses, including SARS-CoV-2 and RSV, or a rapid triplex PCR assay for SARS-CoV-2, RSV, and influenza viruses; the former was generally utilized for hospitalized and high-risk adults while the latter was generally used in the outpatient setting.24 The adoption of these platforms, particularly the rapid triplex assay, closely mirrors the increase in outpatient detections (Supplementary Fig. 3). Altogether, our patient and diagnostic data suggest that the transition to multiplexed platforms for the detection of SARS-CoV-2 led to an incidental increase in RSV detections, particularly among patients not requiring inpatient care. As such, shifts in diagnostic testing platforms must be considered as an additional factor driving case counts in addition to virological and immunological behavior.
RSV Subtype A Infection is Associated with Increased Risk of ICU Admission in Hospitalized Adults
When controlling for clinical and demographic confounders, RSV-A has historically been associated with increased severity and worse patient outcomes compared to RSV-B.5,24,9,10, However, most of these studies were limited to pediatric populations before the COVID-19 pandemic. To determine whether subtype is associated with clinical outcomes in our adult cohort, we subsampled our dataset to only include adult encounters with RSV subtyping information. Of the 3,3416 adult encounters, only 886 were tested on a diagnostic platform that could discern between subtypes, with no typing data available after 2017. To recover subtyping information, we collected 551 residual diagnostic nasopharyngeal swabs from RSV-positive patients between December 18, 2017, and January 1, 2023 (Supplementary Fig. 2). RNA was extracted from each specimen and the RSV subtype was determined using an in-house quantitative reverse transcription PCR (qRT-PCR) assay. Merging these data with our clinical dataset, we recovered typing information for 1,437 patients, including outpatient (n = 495) and inpatient (n = 942) adults.
RSV-A and RSV-B were found to co-circulate each season, with year-to-year shifts in predominance (Fig. 3A). Following the COVID-19 pandemic, subtype predominance fluctuated more dramatically with the 2021–2022 season having a 91.9% RSV-B predominance, and the 2022–2023 season having an 80.8% RSV-A predominance. Inpatient adults with typing information [RSV-A (n = 415) or RSV-B (n = 527)] were highly representative of our overall inpatient cohort (n = 1,565) (Table 1). The age distribution was similar among inpatients with RSV-A or RSV-B, skewing towards 60 years and older (Fig. 3B). Comorbidities were likewise similarly distributed among inpatients with different RSV subtypes, though patients with RSV-B reported slightly lower frequencies overall (Fig. 3C). We applied a Kaplan Meier survival model to observe the probability of discharge among hospitalized patients and found no statistical difference (p-value = 0.620) between subtypes with a 50% discharge probability within 5 days from hospital admission for both A and B infections (Fig. 3D).
To determine whether an association exists between subtype and patient outcome, we used multivariable logistic regression to model subtype infection (i.e., RSV-A versus RSV-B) while controlling for race, ethnicity, sex, age at admission, the month of patient admission, comorbidity sum, hospitalization length of stay, intensive care unit (ICU) admission, and RSV-associated death. This model indicated that ICU admission [adjusted Odds Ratio (aOR) 0.68, 95% confidence interval (CI) 0.49–0.94] and comorbidity sum (aOR 0.92, 95% CI 0.85–0.99) were significantly associated with RSV-A infection (the reference subtype) as compared to RSV-B (Fig. 3E, Supplementary Fig. 4). Likewise, we observed an overall increased proportion of ICU admissions in our RSV-A infected cohort (32.8%) in comparison to the RSV-B cohort (27.1%) (Fig. 3F). These same trends were observed when excluding inpatient encounters after 2020, suggesting that these associations were not confounded by COVID-19 pandemic-associated factors (Supplementary Fig. 4). Altogether, our modeling data are indicative of more severe outcomes in hospitalized adults with RSV-A compared to RSV-B and corroborate previous reports in pediatric populations.
Shift in RSV-B Population Structure Following the COVID-19 Pandemic
The RSV resurgence superseding the low incidence in the 2020–2021 season has been reported to be from previously existing RSV-A lineages that now predominate in the US.25 To determine if the genetic diversity of RSV in Chicago changed throughout the COVID-19 pandemic, we expanded the use of the 551 residual diagnostic nasopharyngeal swabs used for typing for viral whole genome sequencing (WGS). Specifically, viral genomic material from the collected isolates was extracted, reverse transcribed, amplified using a tiled fragment approach spanning the whole genome, and sequenced using the Illumina platform. In total, we obtained 218 whole genome sequences (110 subtype A and 108 subtype B) with a minimum of 90% coverage across the genome. Sequences were given a clade designation using both Nextstrain and G clades nomenclature26 via Nextclade and were deposited in the National Center for Biotechnology Information (NCBI) GenBank (Supplementary Table 1). Finally, maximum-likelihood (ML) phylogenetic trees were generated for all RSV-A and RSV-B sequences.
Between 2017 and 2023, the RSV-A and RSV-B populations in Chicago were exclusively represented by existing clades GA2.3.5 and GB5.0.5a, respectively (Fig. 4A & B). The RSV-A population structure remained largely unchanged after 2020, showing evidence of multiple lineage expansions (Fig. 4A), consistent with prior reports from Arizona27, Washington State28, and Massachusetts25. Conversely, RSV-B exhibited monophyletic clustering with strong support (100% using Shimodaira–Hasegawa approximate likelihood-ratio test [aLRT] and 100% with bootstrap) after 2020 (Fig. 4B), suggestive of population bottlenecking followed by either genetic drift or positive selection. The defining mutations for this monophyletic cluster were mainly in glycoprotein (I252T, I268T, S275P, P214S, P221L, T310I, and S100G) and the fusion protein (S190N, S211N, S389P). To better determine whether this cluster is specific to the Chicagoland area, we pulled 340 publicly available RSV-A and 113 publicly available RSV-B whole genome sequences from the US and repeated our phylogenetic analysis. Again, RSV-A sequences after 2020 were broadly representative of GA.2.3.5 diversity in the US before that year (Fig. 4C). However, RSV-B sequences after 2020 formed a monophyletic cluster composed of our sequences alongside those from Washington and Massachusetts (Fig. 4D). Taken together, these results suggest a potential founder effect in the RSV-B population during the COVID-19 pandemic, which resulted in a more homogenous population structure in the US post-pandemic.
Convergent Evolution of Mutations in the RSV-B Fusion Protein Antigenic Sites
To determine whether this monophyletic cluster was solely identified in the US, we generated a temporal ML phylogenetic analysis using all publicly available RSV-B whole genome sequences collected between 1957 and 2023 (n = 723) (Fig. 5A). This analysis revealed genetically and geographically distinct clusters of RSV-B GB5.0.5a isolates arising in the 2021–2022 season (i.e., unique clusters were observed in Australia, France, Japan, the US, etc.). Similar temporal ML analyses of unique and complete RSV-B G (n = 983) and F (n = 2,067) sequences showed the monophyletic divergent lineage in RSV-B after 2020 (Supplementary Fig. 5).
Subsequently, we used a Bayesian approach to characterize the global population dynamics of RSV-B and compared it with RSV-A (Supplementary Fig. 6) to identify the possible origin and the extent of the expansion of the detected circulating monophyletic lineage in the US. Overall, the population dynamics observed using this approach depicted a temporal evolution of the effective population size for both RSV-A and RSV-B concordant with the epidemiological patterns previously described. More importantly, these analyses confirmed strong statistical support for a monophyletic origin of most of the currently circulating RSV-B viruses in the US with a most likely local origin [ancestral US location probability > 0.99]. Using this approach, we also estimated a time to the most recent common ancestor (TMRCA) for this cluster to be June 9, 2018 [95% Highest Posterior Density (HPD) interval: November 23, 2017 - December 28, 2018] with the next MRCA for this cluster also being identified close to 2 NM sequences sampled between March and April 2018 that constituted a minority lineage at that time. This is indicative of possible selective processes involved in the population bottleneck observed due to the unlikeliness of random drift to a minority lineage instead of within the most prevalent lineages. Although most of the sequences that constituted this cluster were collected in the US, 4 sequences from Austria from NCBI collected during 2022 also clustered within this lineage, indicating ongoing expansion outside the US. On the other hand, RSV-A showed the appearance of multiple unrelated lineages after the period of the SARS-CoV-2 pandemic discarding any possible sampling bias leading to the identification of the dominant monophyletic US cluster in RSV-B.
To further characterize the mutations in the F ORF that define currently circulating isolates and test their possible selective advantage, we generated an ancestral sequence of the most recent common ancestor (MRCA). When comparing the MRCA sequence to all publicly available RSV-B sequences available after 2020 (n = 225), we observed an enrichment of mutations on the heptad-rich A domain (HRA), which is necessary for F protein conformational change and viral-host membrane fusion (Fig. 5B, left).29 When stratifying these data by country, three mutations in the HRA domain (K191R, I206M, and Q209R) were particularly enriched, predominating in 12 of the 15 countries with available sequencing data after 2020 (Fig. 5B, right). While these mutations were all in the majority before 2020, several minority mutations also arose in multiple countries. For example, S190N and S211N, also in the HRA domain, arose to predominance in the US, Austria, and Canada, despite being present in fewer than 2% of sequences before 2020. Not all mutations were localized to the HRA domain, with, for example, P312H and S389P in Domain I which arose in several countries. Notably, most of these mutations lie either in or are flanking antigenic sites Ø, I, and V (Fig. 5C).
Given the position of several of these mutations in or near antigenic regions, we next sought to determine if any of these mutations were under positive selection. Each RSV-B ORF was analyzed independently for episodic diversifying selection using the Mixed Effects Model of Evolution (MEME) method using all available sequence data from 1957 to 2023. Most ORFs showed no evidence of positive selection over this time frame, including NS1, NS2, N, M, P, and SH. M2-1 (T188), M2-2 (S26, H50), and L (S69, S102, N500, I1989) each had positions showing evidence of episodic selection, but these mutations remained low frequency and none are currently circulating (Supplementary Fig. 7). However, both F (Fig. 5D, left) and G (Supplementary Fig. 7) showed high levels of episodic positive selection at positions 7 and 20, respectively. The T250I mutation in G is identified to undergo diversifying selection and is a defining mutation in the monophyletic RSV-B cluster. However, no sites undergoing positive selection in F are fully represented in the cluster, with the F12L mutation only being represented in less than 5% of that cluster. As expected, S173L, a known resistance mutation against Suptavumab in antigenic region V, is now fixed in the RSV-B population. Expanding this analysis to include pervasive positive selection using the Fast Unconstrained Bayesian AppRoximation (FUBAR) method, two sites in the F signal peptide (I5 and F12) are additionally identified (Fig. 5D). If we constrain these analyses to clade GB5.0.5a in the last 5 years, only three mutations are identified to still be under episodic (D548) and pervasive (I5 and F12) positive selection (Fig. 5D, right). While the predominant mutations arising after 2020 are not found to be under statistically significant positive selection, several have high mean posterior difference values between nonsynonymous and synonymous mutations.
Lastly, we quantified the mutational frequency within the binding sites for the monoclonal antibody therapeutics nirsevimab (Fig. 5E), suptavumab, clesrovimab, and palivizumab (Supplementary Fig. 8) before and after 2020 (Fig. 5E). As expected, current RSV-B isolates contain a high frequency of mutations across the Suptavumab binding site both before and after 2020, with treatment-resistant mutations (i.e., S173L) observed in all isolates. However, several mutations in the Nirsevimab binding site, including I206M, Q209R, and S211N, have substantially increased in frequency since 2020 with minor increases observed at other positions (i.e., K68N). Overall, mutations in the Clesrovimab and Palivizumab binding sites remain low.