Ten-Year Surveillance of Respiratory Syncytial Virus Hospitalizations in Adults: Incidence Rates and Case Definition Implications

Abstract

Background

The impact of respiratory syncytial virus (RSV) in older adults is underrecognized, and the limited existing studies on the incidence of hospitalizations show great variability. This study aims to estimate the seasonal incidence rates (IRs) of RSV hospitalizations among adults aged ≥60 years and evaluate how different case definitions influence these estimates.

Methods

A prospective, multicenter observational study with active monitoring was conducted over 10 seasons (2010–2011 to 2019–2020) in 4–10 hospitals (depending on the season) and covered 21%–46% of the region's total population (about 5 million people). RSV hospitalization IRs per 100 000 person-years and 95% confidence intervals were calculated with the exact Poisson method and were stratified by age group (≥60, ≥70, or ≥80 years), RSV season, sex, and the entire study period. Two case definitions were compared: influenzalike illness (ILI) and the combined use of ILI and extended severe acute respiratory infection (ILI/SARI).

Results

A total of 40 600 hospitalizations of individuals aged ≥60 years were included. The RSV hospitalization IRs ranged from 21 to 402 per 100 000 person-years, varying by season, age group, and case definition. The highest IRs were observed in those aged ≥80 years. The ILI case definition underestimated RSV hospitalizations by 13%–40% when compared with the ILI/SARI case definition.

Conclusions

On average, approximately 1 in every 1000 adults aged ≥60 years is hospitalized due to RSV. The risk of a severe RSV infection increases with age and varies significantly between seasons. These are key results for the estimation of the potential impact of the new available RSV vaccines.

The human respiratory syncytial virus (RSV) is a well-known cause of acute respiratory infections in children <5 years of age [1]. However, growing evidence shows a U-shaped age pattern in RSV-related hospitalizations and deaths, with a higher-than-expected burden in older and vulnerable adults [2–10]. One systematic review estimated that the RSV hospitalization rate in individuals aged ≥65 years ranges from 157 to 347 per 100 000, depending on adjustments for underascertainment [11, 12]. However, the reported incidence of RSV hospitalizations in adult populations show considerable variation across studies [2–6, 9, 13, 14].

These differences may stem from variations in case definitions, testing methods, a lack of tailored sampling period, the failure to stratify results by age, and differences in study design, among other factors [13, 15]. In fact, there is still no clear consensus on what case definition is the most suitable for RSV detection in adults [13], hardly allowing for study comparisons. For instance, the World Health Organization's (WHO) definition of influenzalike illness (ILI), which requires the presence of fever, was found to underestimate RSV incidence in older adults by 9-fold [16]. This finding prompted WHO to update its recommendations for hospital-based surveillance for severe RSV infections, to the use of an extended definition of severe acute respiratory infection (“extended SARI”), defined as a hospitalization with an onset of cough within 10 days before admission [17].

With the recent approval of the first vaccines to prevent RSV in adults in the United States and Europe [18–20] and the ones forthcoming [21], there is an urgent need for accurate estimates of the real RSV burden in this population. Well-designed studies that use appropriate case definitions and expanded RSV testing are essential for assessing RSV burden, the public health benefits of RSV vaccines, their effectiveness and their overall impact. In fact, prospective RSV monitoring in hospitalized adults has been identified as a key global health priority by public health experts worldwide [22, 23].

In this study, we leveraged data from the Valencia Hospital Surveillance Network for the Study of Influenza and Other Respiratory Viruses (VAHNSI) [24–27] to estimate the seasonal incidence rates (IRs) of RSV hospitalizations in adults ≥60 years of age from 2010 to 2020. We also address how case definitions influence the IRs estimations.

METHODS

Study Design and Population

A multicenter prospective active-surveillance, observational study within the VAHNSI framework was conducted during 10 seasons (from 2010–2011 to 2019–2020) in 4–10 hospitals (depending on the season; Supplementary Table 1) covering 21%–46% of the total inhabitants of the Valencia region of Spain (about 5 million). Full-dedicated researchers systematically screened and contacted patients of all age groups admitted through the emergency department within the preceding 8–48 hours, with a diagnosis potentially related to respiratory infections. The current study was restricted to patients aged ≥60 years.

Monitoring Period and RSV Season

Monitoring periods included November to March–April (Supplementary Table 1). The RSV season was defined as the period starting from the onset of a sequence of ≥2 consecutive weeks with ≥2 RSV cases, up to the week preceding the initiation of another sequence of ≥2 consecutive weeks without any RSV cases (Supplementary Figure 1). The determination of the start and end of the RSV season was based on RSV cases among ILI cases and across all age groups, including <60 years.

Inclusion Criteria

Noninstitutionalized patients residing within the catchment area of the participating hospitals for ≥6 months, who had no hospital admissions in the prior 30 days and who provided written consent (or had it provided by a representative), were systematically interviewed, and their clinical manifestations and symptoms were recorded. Patients were initially included in the study if, on admission, they exhibited symptoms compatible with ILI, defined as the presence of ≥1 systemic symptom (fever, malaise, myalgia, or headache) and ≥1 respiratory symptom (cough, sore throat, or shortness of breath) with an onset of symptoms within 7 days before admission [28] (Supplementary Table 2). During the 2018–2019 season, patients were also included if they met extended SARI criteria, defined as the presence of cough with an onset of symptoms within 10 days before admission (extended SARI from WHO-SARI [17, 29], referred to hereafter as “SARI” for simplicity). Therefore, only during the 2018–2019 season, patients who met either ILI or SARI criteria (ILI/SARI case definition) were included. RSV testing with multiplex real-time reverse-transcription (RT) polymerase chain reaction (PCR) was performed in all included patients [30].

Case Definitions

To assess the potential underestimation of RSV burden, RSV hospitalizations were estimated using 2 case definitions. First, patients meeting ILI criteria with a laboratory-confirmed RSV infection by RT-PCR were classified as ILI-RSV hospitalizations. Second, given that during 2018–2019 the simultaneous use of ILI or SARI (ILI/SARI) increased the number of included patients with an RSV-positive test (Figure 1), a retrospective analysis of recorded symptoms at admission was conducted across all seasons to identify individuals who would have met the ILI/SARI case definition. These patients were classified as ILI/SARI hospitalizations.

In all seasons, except 2018–2019, patients meeting SARI but not ILI criteria (SARI-not-ILI; Supplementary Table 2) lacked a PCR sample, as they were not included in the study initially. Therefore, we estimated the expected number of ILI/SARI-RSV hospitalizations by assuming that the seasonal RSV positivity rate remained constant between the ILI and ILI/SARI case definitions. This approach allowed us to estimate how many ILI/SARI hospitalizations would have been RSV positive if they had been included in the study since the beginning.

Study Variables

Key variables collected included age, sex, symptoms, and number of days from symptom onset to admission. Variables were collected through interviews with the patients or their representatives.

Sample Collection and Laboratory Analysis

Nasopharyngeal and oropharyngeal samples (FLOQSwabs) were collected from all included patients, combined into a single transport tube, and frozen between −50°C and −20°C until shipped refrigerated to a centralized virology laboratory at FISABIO–Public Health in Valencia. Total nucleic acids were extracted using an automated silica-based method (Nuclisens Easy-Mag; BioMérieux). RSV and other 9 respiratory viruses (and 16 subtypes) were tested by means of multiplex real-time RT-PCR, following WHO protocols with the qScript XLT One-Step RT-qPCR ToughMix (Quanta BioSciences) in a Lightcycler 480II apparatus (Roche Diagnostics). The assay was updated to include new RSV-B circulating clades from the 2014–2015 season onward [31]. Procedures for seasons 2010–2011 and 2011–2012, during which samples were sent to an external laboratory, are described elsewhere [32].

Statistical Analysis

Characteristics of the included patients and those with RSV (sex, age, symptoms, and days from symptom onset to the admission) were described by case definition using frequencies and proportions.

RSV Hospitalization IRs

Crude ILI-RSV and ILI/SARI-RSV IRs were calculated by dividing the total number of ILI-RSV and ILI/SARI-RSV hospitalizations (the numerator) by the person-years (the denominator). The denominator (person-years) for each season was calculated by multiplying the population of the catchment area on January 1 (which is well-defined in the region [33]) by the duration of the RSV season in years. The 95% confidence intervals for IRs were calculated using the exact Poisson method. ILI-RSV and ILI/SARI-RSV hospitalization IRs, along with their 95% confidence intervals, were computed by sex, age group, RSV season, and over the whole study period, and then they were multiplied by 100 000.

To analyze seasonal variation, we performed a Poisson regression analysis. First, we fitted a null model (model 1), which assumes a single constant IR throughout the study period without accounting for seasonal effects. Next, we fitted a second model (model 2), which adjusts the IRs by seasons to account for seasonal differences. The 2 models were then compared using a likelihood ratio test to determine if the seasonal variation was statistically significant. A significant P value (P < .05) from this test would suggest that the season-adjusted model (model 2) provides a better fit to the data, while a nonsignificant P value would indicate no significant variability in IRs across seasons.

We estimated an ILI-RSV underestimation correction factor on the IR by dividing the ILI/SARI-RSV IR by the ILI-RSV IR in each season and age group. The average (SD) across all seasons was considered for each age group.

2018–2019 Season

During the 2018–2019 season, ILI/SARI hospitalizations were included in VAHNSI, and PCR results were available for them. We used the data collected from this season to test our assumption of constant RSV positivity rates between the ILI and ILI/SARI case definitions. We also compared the adjusted ILI/SARI-RSV hospitalization IR estimated for this season, as described above, with the IR obtained from the actual number of laboratory-confirmed ILI/SARI-RSV hospitalizations. All analyses were conducted using R Studio software (R version 4.3.1).

Ethical Statement

The study protocol was approved by the Ethics Research Committee of the Dirección General de Salud Pública-Centro Superior de Investigación en Salud Pública. All participants signed written informed consent before their inclusion in the study.

RESULTS

Study Population

Communication was successfully established with 91% (n = 44 276) of the 48 653 hospitalized individuals aged ≥60 years with a potentially respiratory-related infection and meeting initial inclusion criteria (Figure 1). Of these, 95% (n = 42 090) provided written informed consent, and 40 600 met preliminary inclusion criteria (before ascertainment of case definition). Finally, 18 306 patients (45%) met preliminary and case definition criteria and had a valid PCR sample (Figure 1). Of all included patients, 98.4% (n = 18 013) had both nasopharyngeal and oropharyngeal swab samples for PCR detection, 153 (0.84%) had only an oropharyngeal swab sample, and 140 (0.76%) had only a nasopharyngeal swab sample.

Characteristics of the included patients and those hospitalized for RSV are shown in Table 1. Almost half of the included patients were ≥80 years old (49%). Of the 18 306 patients included and tested, 790 (4.3%) were positive for RSV. The median age of patients with RSV was 81 years, and the proportion of female patients (57%) was higher than all hospitalized patients (45%). The most common symptom among both the included patients and those with RSV was shortness of breath (91% and 94%, respectively), followed by cough (86% and 95%, respectively) and malaise (77% and 73%, respectively). Most of the included patients (87%) were hospitalized within 0–5 days after the onset of symptoms. Supplementary Table 3 shows the characteristics, symptoms, and the duration from symptom onset to hospital admission for each case definition.

Impact of Case Definition

Of the 40 600 patients who met the preliminary inclusion criteria during the entire study period before ascertainment for case definition, 24 583 met the ILI/SARI case definitions, resulting in an estimated 1057 ILI/SARI-RSV hospitalizations across all studied seasons. The ILI case definition led to the detection of 767 RSV hospitalizations (Table 2). The incorporation of the SARI case definition to the existing ILI case definition resulted in an increase of 13%–40% in the estimated number of RSV hospitalizations, depending on the season (Table 2).

RSV Hospitalization IRs

Variations were found among monitoring period durations (from 18 to 47 weeks) (Supplementary Table 1). After the monitoring period was adjusted to the RSV season (with durations from 12 to 26 weeks), 4 RSV cases were excluded: 3 during the 2010–2011 and 1 during the 2018–2019 season (Supplementary Figure 1).

Substantial variability was observed in RSV hospitalization IRs across seasons and age groups, regardless of the case definition (Figure 2 and Supplementary Figures 2 and 3).

RSV hospitalization IRs were similar between female and male patients throughout the seasons (Supplementary Figure 4).

ILI/SARI-RSV hospitalization IRs in the overall study population ranged between 30.3 (2013–2014) and 154.3 (2016–2017) per 100 000 person-years (Figure 2). A likelihood ratio test revealed that the RSV hospitalization IRs varied significantly between seasons (P < 0.01; Supplementary Figure 3). IRs increased with age regardless of the season, and the highest IR was always observed in adults ≥80 years of age, ranging between 44.4 (2013–2014) and 406.3 (2016–2017) hospitalizations per 100 000 person-years. ILI-RSV hospitalization IRs in patients ≥60 years of age ranged between 21.1 (2013–2014) and 112.9 (2018–2019) per 100 000 person-years. In those aged ≥80 year IRs ranged between 27.7 (2013–2014) and 300.9 (2018–2019). The average (SD) ILI-RSV IR underestimation correction factor was 1.4 (0.15) in the overall study population, 1.4 (0.17) in patients aged ≥70 years, and 1.5 (0.2) in those aged ≥80 years.

Season Scenario

During season 2018–2019, the change of ILI to ILI/SARI case definition resulted in an increase from 2858 to 3274 included patients (Supplementary Table 4), of whom 162 were positive for RSV, leading to an RSV positivity rate of 4.9%. Among the ILI hospitalizations the positivity rate was identical: 139 RSV positive among 2858 included patients with ILI, for an RSV positivity rate of 4.9%.

The laboratory-confirmed ILI/SARI-RSV hospitalization IR in the 2018–2019 season was 132 per 100 000 person-years, while the adjusted ILI/SARI-RSV hospitalization IR based on retrospective symptoms search during the same season was 129.2 per 100 000 person-years (Figure 2 and Supplementary Table 4).

DISCUSSION

In this 10-year study of 40 600 respiratory infection–related hospitalizations in individuals aged ≥60 years, we observed important variations in the incidence of RSV hospitalizations, ranging from 21 to 406 per 100 000 person-years, depending on the season, age, and case definition. The risk of hospitalization for RSV increased with age and varied significantly across seasons. Our study highlights the underestimation of RSV burden when using the ILI case definition. This is the first study to estimate the incidence of RSV hospitalizations in older adults in a setting with well-defined catchment areas and dedicated hospital staff, ensuring reliable and comparable results over a decade.

Initially, we used an ILI case definition that included fever as a possible symptom but not a requirement. Since the use of fever as a criterion leads to underestimation of RSV in older adults [13, 29], and ILI alone has been shown to be suboptimal for RSV detection [16, 34], we later adopted a combined ILI/SARI definition. This broader criteria extended the time from symptom onset to hospitalization (from 7 to 10 days) and reduced required symptoms to just cough [29]. In fact, the WHO recommends this SARI definition for monitoring RSV hospitalizations, as it is more sensitive in adults [13]. Similarly, we found that relying on ILI alone underreported RSV cases by 13%–40% depending on the season, with ILI IRs among adults aged ≥60 years underestimated by a factor of 1.4 compared with ILI/SARI.

Our adjusted ILI/SARI IR was estimated based on the assumption that RSV positivity rates remained constant between the ILI and ILI/SARI case definitions. This assumption was corroborated during the 2018–2019 season, when all included patients complying with ≥1 of both definitions were tested with RT-PCR for RSV. Other national studies that use SARI as the main case definition show RSV positivity rates in recent seasons similar to those we observed during 2018–2019 [35]. During that season, the ILI/SARI-RSV hospitalization IRs estimated with both the confirmed number and the estimated number of RSV hospitalizations, based on a retrospective search of symptoms, were very similar (132 vs 129 per 100 000 person-years, respectively). However, we observed that a small percentage of hospitalizations due to RSV met the ILI but not the SARI criteria (10%). Those without cough (a requirement in the SARI definition) would have been excluded had we used the SARI definition only (Supplementary Table 2). We observed that the simultaneous use of ILI and SARI (ILI/SARI) captures the vast majority of RSV cases. Thus, when resources do not allow for the simultaneous use of both inclusion criteria, the use of extended SARI would be preferable. Besides, when the only data available derive from an already set up surveillance network for ILI (eg, from an established influenza sentinel site), an IR adjustment with the estimated underestimation factor could be applied to correct for the ILI underestimation.

Regardless of the case definition used, we consistently observed that hospitalization rates increased with age. The stratification of overlapping age groups (≥60, ≥70, and ≥80 years) was chosen to better capture the RSV burden in populations potentially eligible for vaccination programs, rather than using mutually exclusive age categories. However, overlapping age groups can limit the precision of quantitative comparisons. To address this limitation, we estimated hospitalization rates among patients with ILI in the following age categories: 60–64, 65–69, 70–74, 75–79, and ≥80 years (Supplementary Table 5). The observed rates further support the finding that severe RSV infections become more frequent with advancing age.

Currently published incidence and prevalence rates vary widely among studies [36]. Two studies in the United States and Canada used an active monitoring prospective design similar to ours [37–39]. They also showed substantial variability among age groups and seasons, ranging from 57–666 RSV hospitalizations per 100 000 adults aged ≥60 years (2–4). In industrialized countries, outcomes from the RESCEU project estimated a hospitalization rate in adults aged ≥65 years of about 157 per 100 000 person-years [12]. These rates fall within a range similar to those reported here. However, direct comparisons are difficult due to nuances in study designs, including variations in age group stratifications, misalignment of sampling periods, sample sizes, case definitions or the accuracy of the catchment population used for incidence estimates. Moreover, country-specific healthcare-seeking behaviors can introduce variation in IRs between countries.

There might be an underreporting depending on the sensitivity of RSV detection. A study reported a sensitivity of 93% when using multiplex PCR compared with single-plex RSV PCR [40]. Beyond test accuracy, the type of sample collected significantly affects results. Most studies showed their RSV estimates based solely on PCR testing of nasopharyngeal swab samples, which produces lower detection rates than when combined with sputum samples (52% increase), serology (44% increase), or saliva or mouth/throat swab samples (28% increase) [40, 41]. Consequently, some experts recommend adjusting results for the varied RSV diagnostic testing characteristics related to clinical specimens and testing approaches, by up to 2.2 fold, to account for underestimation in studies relying solely on RT-PCR of nasopharyngeal swab samples[11, 41] (our rates could reach a range of 46–884 per 100 000 person-years in the case of applying such a correction). In our monitoring system, we collected both nasopharyngeal and oropharyngeal samples from most patients (98%) studied, thus minimizing potential underreporting.

Another potential source for bias is a sampling period not tailored to the RSV season [15]. Here, despite using a hospital network initially created for the study of influenza, the monitoring period always started in early autumn, so we probably did not miss a significant number of RSV hospitalizations, as observed in the weekly circulation of RSV cases in our study (Supplementary Figure 1). Moreover, the fact that we restricted the analysis of IR to the RSV season strengthens the results. In addition, we showed no impact on the number of RSV cases in 2 seasons with an extended monitoring period throughout most of the year (seasons 2017–2018 and 2018–2019; Supplementary Figure 1). Other potential limitations in RSV detection among older adults (such as sample size and low case numbers or testing methods) have been described elsewhere [15].

Our study has a few important limitations. First, data from a 4–10-hospital network in a single region may not be representative of the entire country. Even though we conducted active monitoring for the entire RSV season with full-dedicated personnel ascertaining potential cases, 14% of hospitalizations compatible with a respiratory infection were lost due to communication problems or because the patients did not consent to participate (Figure 1). Furthermore, as the study was carried out from Monday to Saturday, screening patients admitted in the previous 8–48 hours, patients admitted between 0 and 8 Am on Saturdays were not screened, and some RSV cases may have been undetected. In addition, we could not estimate the entire catchment area hospitalization rate because older and fragile individuals residing in long-term care facilities were excluded, with our primary focus on community-acquired infections. Nevertheless, internal retrospective assessment revealed that cases not meeting this specific inclusion criterion constituted an average of about 7% (data not shown).

Second, during the first 2 seasons studied (2010–2011 and 2011–2012), the multiplex PCR was performed by a different laboratory, and potential differences resulting from varying testing procedures may have introduced differences in the estimates in these 2 seasons. Third, although both nasopharyngeal and oropharyngeal samples were collected from most patients, we cannot rule out the possibility that the real burden of RSV is underestimated in our series due to sampling, as has been suggested by other authors [11]. Finally, RSV can also cause cardiovascular events or aggravate existing cardiovascular disease. Because during our inclusion procedure we did not screen admitted patients with cardiovascular events, we are likely underestimating the real burden of RSV.

In conclusion, this study provides estimations of the incidence of community-acquired RSV hospitalizations in older adults and the impact of case definitions using an active-monitoring hospital network for respiratory infections. The estimation over the whole study period of the IR of ILI/SARI RSV hospitalizations in people aged ≥60 years ranged between 30.3 and 154.3 per 100 000 person-years. This means that, on average, approximately 1 in every 1000 adults aged ≥60 years is hospitalized due to RSV. The risk of a severe RSV infection increases with age and varies significantly between seasons. These are key results for estimating the potential impact of the newly available RSV vaccines.

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 gratefully acknowledge the staff of the following hospitals in Spain for their support and contribution to the VAHNSI network: Hospital General Universitario de Castellón (Castellón), Hospital Universitario de La Plana (Villarreal), Hospital Universitario y Politécnico La Fe and Hospital Universitario Doctor Peset (Valencia), Hospital Universitario de La Ribera (Alzira), Hospital Lluís Alcanyís de Xàtiva (Xàtiva), Hospital Arnau de Vilanova (Valencia), Hospital Universitario San Juan de Alicante (San Juan de Alicante), Hospital General Universitario de Elda (Elda), Hospital General Universitario de Alicante (Alicante), and Hospital Universitario del Vinalopó (Elche). We also thank all study participants and their families.

VAHNSI members. Members include Mario Carballido Fernández, Maria Dolores Tirado Balaguer, Juan Mollar Maseres, Jose Luis López Hontangas, Maria Dolores Gómez Ruiz, Ana Pineda Caplliure, Juan Alberola, Jose Miguel Nogueira, Juanjo Camarena, Maruan Shalabi Benavent, Francisco José Arjona, Miguel Tortajada-Girbés, Germán Schwarz-Chavarri, Joan Puig-Barberà, Carlos Fernández-García, Laura Cano Pérez, Sandra García Esteban, and Noelia Rodriguez-Blanco.

Author contributions. All authors contributed to the study conception and design. A. U. F. prepared materials, collected and analyzed data, and wrote the first draft of the manuscript. All authors commented on previous versions of the manuscript and read and approved the final manuscript.

Disclaimer. The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.

Financial support. This work was supported by the Chair of Catholic University of Valencia and Moderna; the Instituto de Salud Carlos III (from the European funds of the Recovery, Transformation and Resilience Plan, file code CD22/00122, by virtue of the Resolution of the Directorate of the Instituto de Salud Carlos III, O.A., M.P., of 14 December 2022, awarding the Sara Borrell contracts of the 2022 call of the Strategic Action in Health 2021–2023); and the European Union—NextGenerationEU.

References

Author notes