Estimating the burden of RSV- and influenza-associated ED visits, hospitalizations, ICU admissions, and deaths across age and socioeconomic groups in New York State, 2005–2019

Abstract

Background

Respiratory syncytial virus (RSV) and influenza are leading causes of severe respiratory illness. With multiple RSV prophylactics now available for different age groups, we aimed to assess the pre-intervention burden of RSV across severity levels and risk groups and contextualize the estimates by comparison with influenza.

Methods

We obtained monthly time series data on emergency department (ED) visits, hospitalizations, intensive care unit (ICU) admissions, and deaths by age group, ZIP code, and cause for New York state from 2005 to 2019. Socioeconomic status (SES) of the ZIP codes was classified using supervised principal component analysis (PCA). We estimated the incidence of events attributable to RSV and to influenza using hierarchical Bayesian regression models. Additionally, we assessed severity, defined by ICU admission and mortality risks, as well as the recording fraction (i.e., percent of estimated true virus-associated outcomes recorded as being due to the specific virus), stratified by age, SES, and over time.

Results

The estimated annual incidence of RSV-associated ED visits, hospitalizations, and ICU admissions was highest in infants under 1 in the low SES group (8,100 [95% credible interval (CrI): 7,900-8,200] ED visits, 2,240 [95% CrI: 2,200-2,290] hospitalizations and 330 [95% CrI: 320–350] ICU admissions per 100,000 person-years). The incidence of RSV-associated deaths was highest among adults aged 85 years old and above (61 [95% CrI: 49–74] per 100,000 person-years). In contrast to RSV, the burden of influenza was greatest in age groups 65 years and above. The risk of ICU admission varied by patients’ age and SES, and the mortality risk increased steeply with age for both pathogens from < 2% among infants under 1 to > 10% for the oldest age group (RSV: 11.9% [95% CrI: 9.6–14.3%], influenza: 14.4% [95% CrI: 13.1–15.6%] among age groups 85 year age and above). Incidence varied by epidemic year and season, and we observed an increasing recording fraction of RSV among all age groups over the study period.

Conclusions

RSV and influenza contribute significantly to the burden of ED visits, hospitalizations, ICU admissions, and deaths, particularly among infants and older adults. Although the recording fraction of RSV increased over the study period, it remains lower, particularly for adults. Our findings reveal a substantial disparity in RSV ED visits and hospitalization burden by SES, particularly among younger age groups.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-026-12574-6.

Introduction

Respiratory illnesses caused by viral infections impose a substantial burden of hospitalizations, ICU admissions, and mortality [1]. Respiratory syncytial virus (RSV) and influenza are among the most common pathogens causing lower respiratory tract infections (LRTI), with a particularly high burden in infants, young children, and the elderly [1–3]. Globally, RSV is estimated to cause approximately 3.6 million acute LRTI hospitalizations and 26,300 in-hospital deaths among children aged 0–60 months annually [2]. Comparably, the World Health Organization estimated that globally 3–5 million cases of severe illness and 290,000-650,000 deaths are caused by seasonal influenza per year [4]. Underlying health conditions (e.g., prematurity, chronic lung diseases, congenital heart diseases, compromised immune system) and socioeconomic conditions could increase susceptibility to RSV and influenza infection and severe disease [4–7].

Several prophylactic products protecting against RSV have been approved in the US since 2023, including three vaccines for adults aged 60 years and older (one of which was also approved for pregnant women to protect their newborns) [8, 9] and a long-acting monoclonal antibody for infants [10]. Evaluating the potential impact of these novel preventive interventions requires understanding the baseline burden of disease in different risk groups and for different outcomes.

Estimation of the burden of RSV and influenza is complicated by patterns of healthcare seeking, diagnostic practices, and reporting standards of different healthcare systems. Previous work estimated that the incidence of RSV-associated hospitalizations was highest among infants under 1, followed by adults 85 years of age and older, and was higher among individuals from low-income ZIP codes [11], while the burden of Influenza-associated hospitalizations was estimated to be highest among older adults [12]. In terms of more severe outcomes, i.e., intensive care unit (ICU) admission and death, higher incidence was found among older adults for both pathogens [3, 13–15]. The burden in outpatient settings has been rarely reported. Given the variation in RSV and influenza burden across different clinical outcomes, age groups, and SES groups, a more comprehensive assessment is needed.

This study aimed to quantify the annual incidence of emergency department (ED) visits, hospitalizations, ICU admissions, and deaths due to RSV infection across different age and socioeconomic status (SES) groups, and to contextualize the findings by comparing them to the estimated disease burden attributable to influenza. Our estimates of RSV burden for different subgroups can also serve as a pre-intervention baseline to assess the impact of recently approved prophylactic interventions.

Methods

Data sources

Individual-level inpatient hospital discharge data prior to the COVID-19 pandemic from July 2005 to June 2019 as well as ED discharge data from July 2016 to June 2019 from New York State were obtained from the State Inpatient Databases and State Emergency Department Databases of the Healthcare Cost and Utilization Project (HCUP) [16]. As the coding system of the International Classification of Diseases (ICD) shifted from ICD-9-CM to ICD-10-CM in 2015, data from 2015 were not included in the analysis. We processed and analyzed inpatient (i.e., hospitalization, ICU admission, inpatient-death) and outpatient datasets (i.e., ED visits) separately as they covered different time ranges. Variables extracted from the inpatient database included age at admission, ZIP code of residence, ICD-9-CM (2005–2014) or ICD-10-CM (2016–2019) diagnostic codes, and a binary indicator for death during hospitalization (Table S1). We identified ICU admissions by reviewing individual charge data and marking those with ICU-related charges as individuals admitted to ICU, specifically those with UB-92 revenue center codes 200–219 [17]. We used the ICD-9-CM codes 079.6, 466.11, 480.1 and ICD-10-CM codes B97.4, J21.0, J12.1, J20.5 to identify RSV-coded outcomes, and ICD-9-CM codes 487 and ICD-10-CM codes J09, J10, J11 to identify influenza-coded outcomes and ICD-9-CM codes 460–519 and ICD-10-CM codes J00-J99 to identify all-cause respiratory outcomes.

The population was stratified into nine age groups: under 1, 1, 2–4, 5–9, 10–19, 20–44, 45–64, 65–84, and 85 years old and above. Population size estimates for New York state by age and ZIP code were obtained from the CDC WONDER online database [18]. We classified ZIP codes into three SES groups (low, medium, and high) using supervised principal component analysis (PCA) in which the incidence of RSV-coded hospitalization at the ZIP code level was the target variable, in order to identify the factors most important for explaining variation in RSV incidence (see supplementary methods, Figures S1-3) [19]. The first principal component was derived from seven economic indicators at the ZIP code level, including household income, employment rate, household size, population density, and racial component. We used the first principal component, which accounted for the largest proportion (49%) of the variance to categorize all the ZIP codes into three equal-sized SES tertiles. Patients in our dataset were assigned to one of the three SES groups based on their recorded residential ZIP code, as individual-level indicators of SES were not available.

The individual-level data were aggregated into monthly time series by age and SES group, providing for each month the count of all-cause respiratory ED visits, hospitalizations, ICU admissions, and deaths. We also created time series for RSV-coded hospitalizations under 2 years old by SES and influenza-coded hospitalizations among all age groups by SES, which we used as indicators of RSV and influenza infections over time in the model.

Statistical model

We used hierarchical Bayesian Poisson regression models to estimate the incidence of four all-cause respiratory endpoints (i.e., ED visits, hospitalizations, ICU admissions, and deaths) that could be attributed to RSV and influenza. The analyses were stratified by age and SES groups. The expected number of all-cause respiratory endpoints in age group and SES group in month , , was modeled as a function of indicator for RSV activity, defined as RSV-coded hospitalizations in under 2 years old, and indicator for influenza activity, defined as influenza-coded hospitalizations in all ages, seasonal variation varied by month and temporal trends across seasons :

The effects of each predictor were assumed to be additive. We assume the observed all-cause respiratory endpoints , follow Poisson distributions such that:

The number of hospitalizations, ICU admissions, and in-hospital deaths were linked in the model through disease progression risks and , with the rationale that in-hospital ICU admissions and deaths are subsets of hospitalizations.

We used weakly informative prior distributions for the parameters in our model. A full description of the model and prior structure can be found in the supplementary methods.

Sensitivity analysis

We conducted several sensitivity analyses and compared model outputs to our main results. First, we adopted two alternative model structures: (1) separating data to two periods before and after the ICD transition in 2015 and added period-specific auto-regressive structure for seasonal terms, (2) assuming that observed all-cause respiratory hospitalizations and ICU admissions follow negative binomial distribution. Second, instead of using incidence under 2 as RSV indicator in the main analysis, we alternatively adopted an age-specific indicator for RSV. We used aggregate RSV incidence among three age groups (under 5, 5–44, and 45 years old and above) as indicators of RSV activity in the corresponding age groups within each age range. Third, we used an alternative definition of SES defined based on Social Vulnerability Index (SVI) (details in Supplementary material).

Summarizing model outputs

The incidence of specific outcomes (ED visits, hospitalizations, ICU admissions, deaths) associated with RSV and influenza infection was calculated by multiplying each virus-coded outcome by its estimated coefficient (i.e., scaling factor ) and then dividing by the population size of the specific risk group, i.e., age and SES group. The percent of the all-cause respiratory outcomes that could be attributed to RSV and influenza infection was calculated as the estimated number of virus-associated outcomes divided by the estimated number of all-cause respiratory cases of the specific risk group. We defined the recording fraction of virus-associated hospitalizations as the percent of estimated virus-associated hospitalizations recorded as being due to the specific virus. This was calculated as the ratio of the number of virus-coded hospitalizations recorded in the HCUP database and the estimated number of virus-associated hospitalizations from the model. Disease progression risks (i.e., ICU admission risk, defined as the ratio of ICU admissions and hospitalizations, and mortality risk, calculated as the ratio of deaths and hospitalizations) were also estimated from the model.

Cleaning of the HCUP data was performed with SAS software, version 9.2 (SAS Institute, Cary, North Carolina), and R version 4.3.1 [20, 21]. Statistical analyses and model diagnostics were performed in R version 4.3.1 [21] (Figure S18-24). The model code is available on GitHub (https://github.com/RsvModeling/rsv_burden_deaths_us).

Results

Recorded ED visits, hospitalizations, ICU admissions, and deaths

During the study period, an annual average of 8,100 ED visits, 6,200 hospitalizations, 1,100 ICU admissions, and 63 deaths were recorded as being due to RSV. Among these RSV-coded events, 46.1% of ED visits, 55.4% of hospitalizations, 56.1% of ICU admissions, and 5.4% of deaths occurred in children aged under 1 year old, and 3.3% of ED visits, 10.9% of hospitalizations, 9.8% of ICU admissions and 63.0% of deaths occurred in people 65 years old and above. In contrast, an annual average of 49,000 ED visits, 7,900 hospitalizations, 1,200 ICU admissions, and 250 deaths were recorded as due to influenza. Of these influenza-coded events, 7.1% of ED visits, 44.5% of hospitalizations, 42.3% of ICU admissions, and 69.1% of deaths occurred in people 65 years and above. The annual incidence of RSV-coded hospitalizations and ICU admissions was substantially higher among children under 5 (Figure S4). Older adults had a higher incidence of both RSV- and influenza-coded deaths.

Estimated annual incidence of disease outcomes

Based on our statistical models, the estimated incidence of RSV-associated hospitalizations followed a U-shaped pattern, with the highest rates in the youngest and oldest age groups (Fig. 1; Table S7). It was estimated that 14,200 (95% credible interval (CrI): 13,900 − 14,400) ED visits, 4,600 (95% CrI: 4,500-4,700) hospitalizations, 670 (95% CrI: 650–700) ICU admissions would occur annually among infants under 1 in New York state (Table S7). The estimated incidence rate was highest in infants under 1 year old, with 1,920 (95% CrI: 1,900–1,950) cases per 100,000 person-years, followed by children aged 1-<2 years (600, 95% CrI: 580–610) and adults aged 85 years and above (510, 95% CrI: 480–540). The incidence of RSV-associated ED visits showed a monotonically decreasing trend with age, with a substantial burden estimated for infants and children (Figure S5). It was estimated that 6,100, 4,300, and 2,500 ED visits per 100,000 person-years occurred among children aged under 1, 1 year old, and 2–4 years old.

Fig. 1.

Estimated incidence of RSV- and influenza-associated hospitalizations (A), ICU admissions (B), and deaths (C) in New York state, 2005–2019. The dots indicate the posterior median estimates of the incidence of hospitalizations, ICU admissions, and deaths that are attributable to RSV or influenza per 100,000 person-years, with the labels indicating the estimated values. The error bars indicate the 95% credible intervals of the estimated incidence

The incidence of RSV-associated ICU admissions showed a similar U-shape as hospitalizations, but the incidence of RSV-associated deaths displayed a J-shaped distribution by age (Fig. 1), with a much higher incidence in the age groups 65 years and older (Table S7). Overall, around 18%, 26% and 28% of RSV-associated ED visits, hospitalizations, and ICU admissions occurred in infants under 1, while 72% of RSV-associated deaths occurred in adults aged 65 and above.

Higher incidence of RSV-associated hospitalizations and ICU admissions were observed in the low SES group among children under 10 but not other ages (Fig. 1), while the incidence of RSV-associated ED visits was highest in the low SES group in all age groups (Figure S5). In infants under 1 year old, the incidence rate of RSV hospitalizations was nearly twice as high in the low SES group (2,200 per 100,000 person-years) compared to the high SES group (1,300 per 100,000 person-years), and the incidence of ED visits in low SES group was over three times the incidence in the high SES group (8,100 vs. 2,600 per 100,000 person-years, Figure S5). However, there were no clear differences in the incidence of death from RSV by SES.

In contrast to RSV, incidence rate estimates for all three inpatient influenza-associated outcomes followed a J-shape distribution by age, with the highest burden in the oldest age groups 65 years and older (Fig. 1; Table S7). However, influenza-associated ED visits were highest among infants, similar to RSV (Figure S5). The incidence of influenza-associated ED visits, hospitalization, and ICU admission was markedly lower for children compared with RSV (e.g., 220 [95%CrI: 200–240] for influenza vs. 1,920 [95%CrI: 1,900–1,950] for RSV hospitalizations per 100,000 person-years for infants under 1 year of age, unstratified by SES group, Table S7), while the incidence rate for influenza-associated outcomes was higher than for RSV among older adults 65 years and older. Overall, about 60% of influenza-associated hospitalizations, 58% of influenza-associated ICU admissions, and 82% of influenza-associated deaths occurred in adults aged 65 and above. Models with alternative structures and alternative definitions of SES defined by SVI-composite yielded similar results (Table S4).

Disease burden over time

The burden of disease estimated to be caused by RSV and influenza varied over time (Fig. 2). We estimated a relatively stable trend with a slight decrease in RSV-associated hospitalizations among infants under 1 year old and a decreasing trend in all three inpatient RSV-associated outcomes among adults 65 years and older, in contrast to the increase in inpatient RSV-coded outcomes (Figs. 2, 3 and 4). However, ICU admissions almost doubled in infants under 1 year old between 2013/14 and 2018/19 (Fig. 2). This rise in ICU admissions for infants coincided with shorter hospital stays, increased use of non-invasive mechanical ventilation (NIMV), and decreased use of invasive mechanical ventilation (IMV) (Figures S6, S7). A larger decrease in the length of stay in the hospital also coincided with a larger increase in ICU admissions in the low SES group (Figures S6, S8). In contrast with inpatient RSV-associated burden over time, influenza-associated outcomes remained relatively stable over the study period, albeit with more season-to-season variability for older age groups (Figs. 2, S9, S10). The incidence of ED visits for both pathogens remained stable from 2016 to 2019 (Figure S11).

Fig. 2.

Estimated incidence of RSV- and influenza-associated hospitalizations, ICU admissions, and deaths over time by age group in New York state, 2005–2019. The pale grey blocks represent the time period of transition of the ICD coding system from ICD-9-CM to ICD-10-CM in 2015. The dots represent the posterior median estimates of the incidence rate of hospitalizations, ICU admissions, and deaths that are attributable to RSV or influenza per 100,000 person-years for each season. The error bars indicate the 95% credible intervals of the estimated incidence. The four vertical panels indicate the incidence estimates for age groups under 1, 1-<2, 65–84, and 85 years old and above

Fig. 3.

Monthly incidence of recorded RSV-coded, estimated RSV-associated, and estimated all-cause respiratory outcomes among infants under 1 year old in New York state, July 2005 - June 2019. The red area represents the incidence of outcomes recorded as being due to RSV in the HCUP database (RSV-coded outcomes). The yellow area represents the posterior median incidence of RSV-associated outcomes estimated from the model. The blue area represents the posterior median incidence of all-cause respiratory outcomes estimated from the model

Fig. 4.

Monthly incidence of recorded RSV-coded, estimated RSV-associated, and estimated all-cause respiratory outcomes among adults aged 65 and above in New York state, July 2005 - June 2019 (left: original; right panel: zoomed in). The red area represents the incidence of outcomes recorded as being due to RSV in the HCUP database (RSV-coded outcomes). The yellow area represents the posterior median incidence of RSV-associated outcomes estimated from the model. The blue area represents the posterior median incidence of all-cause respiratory outcomes estimated from the model

Percent of respiratory outcomes attributable to RSV and influenza

RSV infections were estimated to contribute to a large proportion of respiratory ED visits, hospitalizations, ICU admissions, and deaths in children under 5 years old, especially in infants under 1 year old, compared to influenza (Figs. 5, S12). Without stratification by SES, we estimated that 33.9% (95%CI: 33.1–34.4%) of respiratory ED visits, 43.7% (95% CrI: 43.1–44.3%) of respiratory hospitalizations, 33.5% (95% CrI: 32.3–34.8%) of respiratory ICU admissions, and 13.1% (95% CrI: 10.7–15.8%) of respiratory deaths in the first year of life could be attributed to RSV infection. The attributable percentage for all four outcomes was similar across SES groups (Figs. 5, S12) and showed strong seasonality (Figure S13A). During the winter season, up to 60–80% of respiratory outcomes were estimated to be attributed to RSV among young children. For influenza, in contrast, the attributable percentage was low (< 5%) for hospitalizations and ICU admissions. However, it was higher (10–20%) among school children (Fig. 5). The attributable percentage of influenza also exhibited seasonality, with a higher attributable percentage during winter seasons (Figure S13B).

Fig. 5.

Percent of respiratory outcomes attributable to RSV and influenza infection in New York state, 2005–2019. The bars show the posterior median percent of all-cause respiratory outcomes (hospitalizations, ICU admissions, and deaths) that were estimated to be attributable to RSV or influenza infection, calculated as the estimated incidence of RSV-/influenza-associated outcomes divided by the estimated incidence of all-cause respiratory outcomes. The texts above the bars label the detailed numbers of the medians. The error bars indicate the 95% credible intervals of the estimated attributable percent

Inpatient ICU admission and mortality risk

The proportion of RSV- and influenza-associated hospitalizations resulting in ICU admission or death showed distinct patterns by age group (Fig. 6). There was more variation by age in the ICU admission risk for RSV than for influenza (Figure S14A). The percent of RSV-associated hospitalizations admitted to the ICU was highest in the 65–84 year age group (17.5%), followed by the 45–64 year age group (15.6%) and infants under 1 year (14.6%) (Figure S14A). The mortality risk of both RSV- and influenza-associated hospitalizations increased with age from less than 1% in the age groups < 5 years to over 10% in the age groups 65 years old and above (Fig. 6). We estimated a higher risk of ICU admission and death in the higher SES group for both RSV and influenza (Figure S14B), but this disparity persisted across all age groups only for the ICU admission risk of RSV (Fig. 6A). The estimated proportion of hospitalizations admitted to the ICU was generally lower than the observed proportion directly calculated using the virus-coded outcomes for both RSV and influenza (Figure S15).

Fig. 6.

Estimated proportion of RSV- and influenza-associated hospitalizations admitted to the ICU admission (A) and resulting in death (B), 2005–2019, New York state. The dots indicate posterior median estimates of the proportion of RSV- and influenza-associated hospitalizations admitted to the ICU or dying. The error bars indicate the 95% credible intervals of the estimated ICU admission and mortality risk. ICU admission risk is defined as the ratio between the incidence of virus-associated ICU admission and the incidence of virus-associated hospitalizations; the mortality risk is defined as the ratio between the incidence of virus-associated deaths and the incidence of virus-associated hospitalizations

Recording fraction of RSV- and influenza-associated hospitalizations

Across all seasons, the percentage of estimated RSV-associated hospitalizations recorded as such was highest in infants under 1 year old (74.4%) and decreased with age, dropping to around 10% in adults (Fig. 7A). However, over the study period, recording fractions rose in all age groups (Fig. 7B). Among adults 65 years and older, the recording fraction increased from under 5% in the 2005/06 season to around 20% in the 2017/18 season and over 50% in the 2018/19 season.

Fig. 7.

Percent of pathogen-associated hospitalizations recorded as being due to the pathogen. Panel A) and C): Recording fraction of (A) RSV- and (C) influenza-associated hospitalizations, 2005–2019, New York state. The recording fraction is defined as the ratio of the number of pathogen-coded and estimated pathogen-associated hospitalizations. The bars represent the posterior median estimates of the recording fractions, and the error bars represent the 95% credible intervals of the estimates. Panel B) and D): Recording fraction of (B) RSV- and (D) influenza-associated hospitalizations over time by age. The dots represent the median estimates of the recording fraction, and the error bars represent 95% credible intervals of the estimates

Our estimated recording fractions for influenza-associated hospitalizations exceeded 100% especially for children aged 2–4 years old (Fig. 7C), possibly indicating an over-attribution of etiology of respiratory hospitalizations to influenza. Similar to RSV, the lowest recording fraction for influenza was also observed in the older age groups, although it was increasing over the study seasons (Fig. 7D). During the study period, the recording fractions for ED visits were generally low for RSV among all age groups and were higher in higher SES groups (Figure S16).

Discussion

This study provided burden estimates of ED visits, hospitalizations, ICU admissions, and deaths due to RSV and influenza by age and SES group. We also illustrated how the burden evolved over the study period from 2005 to 2019, which ended before RSV and influenza activities were disrupted by the COVID-19 pandemic and before the approval of several prophylactic products against RSV [22–25]. Our results demonstrated that 18%, 26% and 28% of RSV-associated ED visits, hospitalizations, and ICU admissions occurred in infants under 1, while 72% of RSV-associated deaths occurred in the age groups 65 years old and above. Older adults had the highest inpatient influenza burden, with 60%, 58% and 82% of influenza-associated hospitalizations, ICU admissions, and deaths occurring in the age groups 65 years old and above. Although the recording fraction of RSV was found to be increasing for all ages over time, the gap between recorded and estimated RSV hospitalizations was still largest in older adults.

Assuming similar incidence across states, we extrapolate that 1,340,000 and 1,760,000 ED visits, 300,000 and 290,000 hospitalizations, 42,000 and 31,000 ICU admissions, 14,000 and 25,000 deaths due to RSV and influenza respectively would occur in the US every year (Table S5). Among infants < 1 year of age, 230,000 RSV-associated ED visits, 72,000 hospitalizations, 11,000 ICU admissions, and 240 deaths would occur annually. Among adults aged 65 and above, 97,000 RSV-associated hospitalizations, 16,000 ICU admissions, and 10,000 deaths would occur annually. Additionally, 190,000 influenza-associated hospitalizations, 19,000 ICU admissions, and 21,000 deaths would occur in adults aged 65 and above.

Our age-specific incidence estimates for RSV-associated hospitalizations in the US are generally consistent with previous studies [3, 11, 26–28]. Our estimates for older adults before the ICD transition in 2015 are slightly higher than previous estimates but are similar to those estimated using a model fit to the same database [11]. For seasons after 2015, our model provided comparable results to Branche et al., which also used data from New York and involved intensive clinical sampling among adults [27]. We reported similar estimates for death as in previous literature [3, 29]. Suh et al. estimated a nationwide incidence of 0.13 million ED visits per year among infants under one based on data from 2011 to 2019 [30]. After accounting for under-recording, we reported a higher incidence of 0.23 million per year for the similar time frame. Our influenza-associated estimates are also comparable with previous estimates in the literature [15, 31–34], while offering greater granularity by providing estimates by SES and using finer age groups.

We also attempted to compare our estimates with data from the Epic Cosmos database to validate our results (Table S6) [35]. Our estimates of the percentage of respiratory encounters attributable to RSV across age groups generally aligned with the RSV positivity rate among respiratory encounters in Cosmos, though some differences were observed in the inpatient setting (Table S6). The attributable fraction can be thought of as the product of the probability that a patient is tested and the positivity rate. Positivity rate usually depends on characteristics of the tested population for whom the test is ordered, and test-seeking behavior could influence the positivity rate of a certain pathogen if the probability of being tested is associated with different levels of severity, clinical setting, and different possibilities of receiving tests for different pathogens, and the type of test used. In the Cosmos database, RSV testing occurs more frequently in the ED than in inpatient settings (Figure S17), but testing in both settings increased over the seasons, consistent with our findings of the increasing recording fraction. Infants under one year old, particularly those with severe symptoms, might be more likely to be tested during their first ED visits to guide the timely prescription of antiviral treatments for RSV or other respiratory viruses. In Cosmos, RSV testing is less common among other age groups due to the absence of RSV-specific treatments, and those hospitalized had higher positivity rates, reflecting increased disease severity. In contrast, our estimates aimed to capture RSV burden across different severity levels, adjusting for variations in testing practices by severity, which may explain some discrepancies between our estimates and the observed data from Cosmos.

Our study examined RSV and influenza burden across different SES groups classified using supervised PCA that incorporated several socioeconomic variables. A previous study by Zheng et al. classified SES by household income and reported a higher incidence of RSV-associated hospitalizations in the low-income group among all age groups. For ED visits, we found that higher incidence was associated with lower SES in all age groups for both RSV and influenza. This could be due to larger household size and number of contacts, higher prevalence of prematurity and higher rate of underlying conditions like asthma [36, 37]. We found the same trend by SES in the incidence of RSV hospitalizations and ICU admissions only among children under 10, but not in older age groups. There have been mixed findings on the impact of SES on respiratory disease burden among adults [5, 38, 39]. One study conducted in New Zealand found that living in low SES neighborhoods was associated with increased RSV hospitalization rates in adults [40]. Other studies found SES was associated with increased incidence of RSV- and influenza-associated hospitalizations, but not with more severe outcomes such as ICU admission and death [38, 41]. Our model also estimated a higher ICU admission risk and mortality risk in the higher SES group for both pathogens, and this disparity persisted in all age groups in ICU admission risk for RSV (Fig. 6). The ICU admission risk directly calculated from the RSV-coded outcomes showed a similar trend (Figure S12B). This might be due to the disparity in the ICU capacity between hospitals and the impact of a family’s SES on the decision to admit patients into ICUs. Another possible reason is that the threshold for admission to hospital might be lower for low-SES individuals, and this would inflate the denominator for ICU admission risk. Considering the substantial disparity across SES in incidence of RSV-associated hospitalization and ICU among infants, expanding coverage of the currently available monoclonal antibody and maternal RSV vaccine in lower SES groups could help reduce these inequities and provide greater population-level protection.

Our results showed that the proportion of RSV-associated hospitalizations admitted to the ICU had a larger variation by age for RSV, compared to a rather stable trend by age for influenza. Relatively constant ICU admission risks of influenza-associated hospitalizations by age were also estimated by O’Halloran et al. [34]. Previous studies on RSV disease progression have predominantly focused on in-hospital ICU admission and mortality risks, relying on individual-level clinical data. Walsh et al. reported an ICU admission risk of 28% among adults hospitalized due to RSV (2014–2016) and did not observe a significant difference by age [42]. A recent study among US adults reported that 24.8% of RSV hospitalizations were admitted to the ICU and 12.0% required IMV or died [13], which is higher than our direct estimates of 17.2% ICU admission and 5.0% mortality risk among adults calculated directly from the virus-coded outcomes (Figure S12), and higher still than our estimates from the model that accounted for underreporting. This indicates that our model additionally captured less severe hospitalizations that were missed by healthcare facilities or incorrectly coded as due to other causes.

We highlighted an elevated incidence of RSV- and influenza-associated outcomes among adults 65 years and older for both RSV and influenza and found that the inpatient mortality risk was also considerably higher for both pathogens in this age group, potentially due to a much higher prevalence of existing morbidity and complications. Our results again captured the significantly lower recording fraction of RSV hospitalizations in this age group as indicated by Zheng et al. [11]. Furthermore, we observed an increasing trend of recording fractions of RSV in all ages over the study period (2005–2019). At the beginning and during the COVID-19 pandemic (2018–2022), data from Epic Cosmos also showed a substantial increase in RSV testing rates in ED across all age groups (Figure S17). While our recording fraction is not directly comparable to the test rate in Cosmos, the upward trends in both metrics might suggest improved RSV testing and diagnosis over the past decade. Our model also estimated that the recording fraction of influenza was larger than 1 among children aged 1–4 years old, especially during 2009/10 and 2013/14 seasons when influenza activity was high. This may indicate misclassification or over-attribution of other pathogens to influenza without confirmatory testing during these periods among young children.

Our incidence estimates over time showed that RSV-associated ICU admissions in children under 2 years old almost doubled over the studied period, despite a moderate decrease in RSV-associated hospitalizations. A similar trend was also observed in the Netherlands, which reported a fourfold increase in pediatric ICU admissions for RSV bronchiolitis among children under 2 years old from 2003 to 2016 [43]. We found that this increase was accompanied by a decreasing length of in-hospital stay for ICU-admitted RSV cases, an increasing number receiving NIMV, and a decreasing number receiving IMV over the years in this age group (Figure S7, S8). Changes in the options of ventilation support upon ICU admissions might have influenced the ICU admission threshold, leading to less severe cases being admitted to the ICU over more recent seasons. What the admission criteria are, and how they are evolving in clinical practices, is often complicated and requires further study with more detailed data from clinical settings.

Limitations

Our study has several limitations. First, the time series data we used included a one-year gap in 2015 due to the transition from ICD-9 to ICD-10. Although the transition to ICD-10 introduced an additional category for acute bronchitis due to RSV (J20.5), this category was previously encompassed within one of the three ICD-9 codes for RSV and thus was not expected to significantly impact case classification (Figure S18). Nonetheless, caution is warranted when comparing estimates from these two periods. Additionally, the diagnosis of hospitalizations based solely on ICD codes is susceptible to misclassification, particularly since RSV and influenza share similar symptoms with many other respiratory infections.

Second, the classification of ICU admission was not based on clinical records but the ICU-related revenue codes in the medical charge dataset, as they were the only available indicators for ICU admission in the database. Some patients might only stay in the ICU briefly and later be transferred out but were also counted as admitted to the ICU. The gold standard defines ICU admission only if a patient remains in the ICU for at least two consecutive hours or longer [17]. Although studies have shown revenue codes provide high sensitivity and specificity in identifying true ICU admissions [17], our classification might additionally capture some less severe cases.

Third, our finding of the increasing recording fraction relies on the validity of estimates of RSV-associated hospitalizations, which might be hard to validate. Nevertheless, our estimates of hospitalizations are similar with previous studies, and our estimates of attributable percentage among young children align with the directly calculated test rate from Cosmos (Figs. 5, S12; Table S6). Furthermore, we observed an increasing recording fraction in all age groups, including the youngest age group for which estimates were usually more reliable due to more frequent testing. Our model also estimated that the recording fraction of influenza for children aged 2–4 years old was larger than 100%, indicating that respiratory hospitalizations in this age group might have been overly attributed to influenza, thus leading to underestimation of RSV incidence in this age group.

Conclusions

This study provided a comprehensive overview of the burden of RSV and influenza before the COVID-19 pandemic. We found that RSV and influenza disproportionately affected different age groups and exhibited varying impacts across levels of disease severity. Our results showed that infants under 1 year of age had the highest incidence of RSV-associated ED visits, hospitalizations and ICU admissions, while older adults aged 65 and above had the highest incidence of RSV-associated death. Older adults additionally experienced a higher incidence of influenza-associated hospitalizations, ICU admissions, and deaths. Although the largest gap between recorded and estimated RSV hospitalizations was observed in older adults, the recording fraction for all age groups has been increasing over the study period. The proportion of RSV- and influenza-associated hospitalizations admitted to the ICU varied by age, and the mortality risk was considerably higher in older adults.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

RSV

Respiratory syncytial virus

SES

Socioeconomic status

US

The United States

ED

Emergency department

ICU

Intensive care unit

PCA

Principal component analysis

CrI

Credible interval

LRTI

Low respiratory tract infection

HCUP

Healthcare Cost and Utilization Project

ICD

International Classification of Diseases

SVI

Social vulnerability index

NIMV

Non-invasive mechanical ventilation

IMV

Invasive mechanical ventilation

Author contributions

HX implemented the study, analyzed the data, and drafted the article. JLW designed the study’s analytic strategy. EDS helped revise the manuscript. VEP conceptualized the study and consulted on the analyses and revisions of the manuscript. DMW conceptualized and designed the study’s analytic strategy, reviewed and revised the manuscript. SP helped pull and aggregate data from Epic Cosmos. All authors have reviewed and approved the final draft of the manuscript.

Funding

This work was supported in part by the National Institutes of Health (NIH) grant number R01AI137093. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Data availability

The HCUP hospitalization data are not available publicly but can be obtained from the State Inpatient Database upon signing a data use agreement with the Agency for Healthcare Research and Quality. The code used for this project can be found in the Github repository: https://github.com/RsvModeling/rsv_burden_deaths_us.

Declarations

Ethics approval and consent to participate

The analysis of the data was approved by the Human Investigation Commit- tees at Yale University.

Consent for publication

Not applicable.

Competing interests

DMW has been the principal investigator on grants from Pfizer and Merck to Yale University for work unrelated to this manuscript and has received consulting fees from Pfizer, Merck, GSK, and Vaxcyte for work unrelated to this manuscript. JLW has received consulting fees from Pfizer for work unrelated to this project. SP has been the principal investigator on grants from Merck to Yale University for work unrelated to this manuscript and has received travel fees from Pfizer and Merck unrelated to this manuscript.