Estimating the Hospitalization Burden Associated With Influenza and Respiratory Syncytial Virus in New York City, 2003–2011

Edward Goldstein; Sharon K. Greene; Donald R. Olson; William P. Hanage; Marc Lipsitch


Influenza Resp Viruses. 2015;9(5):225-233. 

In This Article

Abstract and Introduction


Background Hospitalization burden associated with influenza and respiratory syncytial virus (RSV) is uncertain due to ambiguity in the inference methodologies employed for its estimation.

Objectives Utilization of a new method to quantitate the above burden.

Methods Weekly hospitalization rates for several principal diagnoses from 2003 to 2011 in New York City by age group were regressed linearly against incidence proxies for the major influenza subtypes and RSV adjusting for temporal trends and seasonal baselines.

Results Average annual rates of influenza-associated respiratory hospitalizations per 100 000 were estimated to be 129 [95% CI (79, 179)] for age <1, 36·3 (21·6, 51·4) for ages 1–4, 10·6 (7·5, 13·7) for ages 5–17, 25·6 (21·3, 29·8) for ages 18–49, 65·5 (54·0, 76·9) for ages 50–64, 125 (105, 147) for ages 65–74, and 288 (244, 331) for ages ≥75. Additionally, influenza had a significant contribution to hospitalization rates with a principal diagnosis of septicemia for ages 5–17 [0·76 (0·1, 1·4)], 18–49 [1·02 (0·3, 1·7)], 50–64 [4·0 (1·7, 6·3)], 65–74 [8·8 (2·2, 15·6)], and ≥75 [38·7 (25·7, 52·9)]. RSV had a significant contribution to the rates of respiratory hospitalizations for age <1 [1900 (1740, 2060)], ages 1–4 [117 (70, 167)], and ≥75 [175 (44, 312)] [including chronic lower respiratory disease, 90 (43, 140)] as well as pneumonia & influenza hospitalizations for ages 18–49 [6·2 (1·1, 11·3)] and circulatory hospitalizations for ages ≥75 [199 (13, 375)].

Conclusions The high burden of RSV hospitalizations among young children and seniors age ≥75 suggests the need for additional control measures such as vaccination to mitigate the impact of annual RSV epidemics. Our estimates for influenza-associated hospitalizations provide further evidence of the burden of morbidity associated with influenza, supporting current guidelines regarding influenza vaccination and antiviral treatment.


Estimating the burden of severe outcomes associated with influenza and respiratory syncytial virus in various population groups is important to inform mitigation efforts, including potential development of an RSV vaccine.[1–4] With most hospitalizations associated with influenza and RSV infections missing a mention of those pathogens in the diagnoses, substantial uncertainty still exists about the magnitude of this burden and the statistical method of estimating it.[5–9] The aim of our work was to both address the statistical issues related to the estimation of the hospitalization burden associated with influenza and RSV and exhibit a detailed picture of this burden by considering a number of principal hospitalization diagnoses that have a potential contribution from influenza and RSV.

While much of the related recent work on estimating the severe outcome burden associated with influenza and RSV relies on regression analysis,[10–15] several aspects of such methodologies are questionable. One is that the commonly employed model assumes a nonlinear relation between exposure (influenza incidence rates) and outcome (influenza-attributable hospitalization)[6,16] while the rates of severe outcomes associated with influenza are expected to be proportional to influenza incidence rates, and the implications of modeling the relation between incidence and severe outcomes in a nonlinear fashion are uncertain.[13] Many estimates also use a sinusoidal model to account for baseline rates of outcomes not associated with respiratory viruses.[6,8,17] This assumption might be more problematic for RSV compared with influenza due to the high year-to-year periodicity of RSV circulation (particularly compared with influenza,[7] which could confound the estimation of RSV-attributable outcomes if the true baseline varies in a fashion correlated with RSV.

The approach proposed in ref.[13] was designed to overcome these problems, employing a linear relation between the rates of influenza incidence and severe outcomes and adopting a flexible model for the baseline rates of severe outcomes not associated with influenza. Here, we apply a similar approach to the New York City data between 2003 and 2011, adding an appropriate RSV incidence proxy to the inference model in ref.[13] This inference method is applied to data on respiratory hospitalizations [including hospitalization with principal diagnoses of pneumonia and influenza (P&I) and chronic lower respiratory disease (CLRD)] as well as for hospitalization with a number of other principal diagnoses (including circulatory causes) in different age groups in New York City between 2003 and 2011 to estimate the burden for the above categories of hospitalization associated with influenza and RSV.