Population-Based Surveillance for Hospitalizations Associated With Respiratory Syncytial Virus, Influenza Virus, and Parainfluenza Viruses Among Young Children
Objective. Respiratory syncytial virus (RSV), influenza virus, and parainfluenza viruses (PIV) cause significant morbidity in young children. Although only influenza virus infection and illness is currently vaccine-preventable, vaccines are under development for RSV and PIV. We established a prospective, active population-based surveillance network to provide precise estimates of hospitalization rates for viral acute respiratory illness (ARI) in young children and to measure the potential impact of enhanced vaccine usage on these rates.
Methods. Prospective, active population-based surveillance was conducted in young children who were hospitalized for ARI from October 1, 2000, to September 30, 2001, in Monroe County, New York (Rochester area) and Davidson County, Tennessee (Nashville area). Eligible children younger than 5 years were those who resided in surveillance counties and were hospitalized for febrile or acute respiratory illness. Viral culture and polymerase chain reaction identified viruses from nasal and throat samples obtained from all surveillance children. We measured population-based rates of hospitalization for RSV, influenza virus, and PIV as well as demographic, clinical, and risk factor assessment for each virus.
Results. Of 812 eligible hospital admissions, 592 (73%) children were enrolled. Of the enrolled children, RSV was identified in 20%, influenza in 3%, PIV in 7%, other respiratory viruses in 36%, and no detectable virus in 39%. Population-based rates of ARI hospitalizations in children younger than 5 years were 18 per 1000. Virus-positive hospitalization rates per 1000 children were 3.5 for RSV, 1.2 for PIV, and 0.6 for influenza virus. Younger age (particularly <1 year), black and Hispanic race/ethnicity, male gender, and presence of chronic underlying illness were associated with higher hospitalization rates.
Conclusions. This study confirms that children younger than 5 years and particularly children younger than 1 year have a high burden of hospitalization from RSV, influenza, and PIV. The enhanced use of influenza vaccine and the development of RSV and PIV vaccines have the potential to reduce markedly the pediatric morbidity from ARIs.
Acute respiratory illnesses (ARIs) cause significant morbidity in young children.1–6 Although earlier studies have assessed the pediatric burden of ARIs, most were not prospective, population-based studies and did not specifically target potentially vaccine-preventable viruses. The expanded recommendations for inactivated influenza vaccine7 in young children, the recent licensure of the live attenuated intranasal influenza vaccine,8 and ongoing efforts at vaccine development for respiratory syncytial virus (RSV) and parainfluenza virus (PIV)9,10 make the establishment of such an accurate surveillance system timely and important. In addition, the recent development and use of polymerase chain reaction (PCR) methods have markedly improved the etiologic identification of specific viral pathogens.11,12
The New Vaccine Surveillance Network (NVSN) of the Centers for Disease Control and Prevention (CDC) was recently established to provide a precise estimate of the burden of ARI from vaccine-preventable or potentially preventable agents and to monitor the impact of vaccine use on disease rates. Prospective, population-based surveillance of children who were younger than 5 years, resided in 2 urban counties in New York and Tennessee, and were hospitalized with fever or ARIs was conducted throughout the year. PCR methods and viral culture of nasal and throat swab samples were used to identify specific respiratory viruses. This report presents results from the first year of surveillance and clearly demonstrates the potential of the NVSN.
Population-based rates of hospitalizations associated with laboratory-confirmed acute respiratory viral illness in children who were younger than 5 years and residing in Monroe County, New York, and Davidson County, Tennessee, were determined. Influenza virus, RSV, and PIV were the focus of this report given their known contribution to serious respiratory illness in children and because of the NVSN focus on infections for which vaccines are available or under development. The study of human metapneumovirus has been presented elsewhere.13,14 Surveillance was conducted from October 1, 2000, through September 30, 2001, in 2 hospitals in Monroe County and 3 hospitals in Davidson County, which admitted >95% of the hospitalized children who reside in those counties (see below). The hospitals are located in the cities of Rochester, New York, and Nashville, Tennessee, which are the urban centers of the counties. Children in the surrounding suburban and rural areas in each county are also admitted to these hospitals; children from outlying counties were excluded because a greater proportion were hospitalized elsewhere. Institutional Review Boards of the 2 study sites, the participating surveillance hospitals, and the CDC approved the study.
Study nurses identified children who were hospitalized between 7 am on Sunday and 7 am on Thursday at each surveillance hospital each week. Children were considered for enrollment when they were younger than 5 years, resided in Monroe or Davidson counties, were admitted within 48 hours of an enrollment period, and had an admitting diagnosis (as reported by the admitting physician) suggestive of an ARI or ARI-related condition (acute respiratory illness, apnea, asthma, bronchiolitis, croup, cystic fibrosis exacerbation, febrile neonate, febrile seizure, fever without localizing signs, hypothermia, influenza, otitis media, other respiratory infection, paroxysmal cough, pharyngitis, pneumonia, respiratory distress, RSV, rule-out sepsis, sinusitis, group A β-hemolytic streptococcal pharyngitis, tonsillitis, upper respiratory illness, wheezing). Furthermore, children were excluded when they had a known nonrespiratory cause for hospitalization, had respiratory symptom duration >14 days, were transferred from another surveillance hospital where they would have been already enrolled, were discharged from a hospital in the previous 4 days, or were newborns who had never left the hospital. When eligibility was unclear to study nurses, they referred the case to a study investigator, who made the final determination on the basis of information available at admission. Informed written consent was obtained from each child’s parent or guardian before enrollment.
Nasal and throat swabs were obtained from each enrolled child, combined into a tube of veal-infusion broth transport medium, and delivered at ambient temperature within 1 to 2 hours to the research virology laboratory at each site. Viral cultures were processed immediately. For reverse transcription-PCR (RT-PCR) testing, sample aliquots were collected in lysis buffer and frozen at −70°C until shipped in batches to the CDC for RNA extraction and RT-PCR. The details of the laboratory methods have been described previously.11,12 RSV, influenza virus types A and B, and PIV types 1 to 3 were the primary virus types targeted for detection by tissue culture methods at research laboratories in Rochester and Nashville. Aliquots of each specimen were inoculated onto conventional viral cell cultures using primary rhesus monkey kidney, HEp-2, and NCI-H292 cells and incubated at 35°C. Viruses were initially identified by cytopathic effect and hemadsorption. Cultures without cytopathic effect or hemadsorption at 10 days were screened further by immunofluorescent assay. The CDC laboratory performed RT-PCR to detect RSV, influenza virus types A and B, PIV types 1 to 3, human metapneumovirus, and picornaviruses (not further characterized as enterovirus or rhinovirus). RT-PCR assays were designed using fluorescence-labeled primers and automated capillary electrophoresis and GeneScan DNA fragment analysis for amplicon detection. Specimens were designated PCR positive when the same amplification product was found in a second independent RT-PCR using a duplicate specimen aliquot. As previously described,11 the RT-PCR procedures were highly reproducible, sensitive, and specific. A specimen was defined as virus positive when 1 viral culture or 2 independent RT-PCR tests were positive.
Demographic and Clinical Information
Demographic information, symptom duration, history of underlying medical conditions, exposure to tobacco smoke, number of children in the household, breastfeeding, child care attendance, influenza vaccination status, palivizumab administration, and antiviral medications were collected from parent/guardian interview. History of underlying medical conditions, health insurance status, admission diagnoses, microbiology and chest radiograph results, hospital course, and discharge diagnoses were obtained from medical chart review. A maximum of 10 International Classification of Diseases, Ninth Revision discharge diagnosis codes were recorded for each hospitalization, on the basis of diagnoses recorded and coded by nonstudy hospital coding staff. Medical conditions for which influenza vaccine is routinely recommended (heart disease, chronic pulmonary disease, kidney disease, cancer, diabetes, immunodeficiency, and sickle cell anemia) were included.
Population-Based Hospitalization Rates
Hospitalizations per 1000 children were expressed as the weighted number of ARI hospitalizations divided by the number of children in the county population determined by the 2000 US Census, multiplied by 1000. Rates were calculated using SAS software15 procedures by weighting the observed number of enrolled hospitalizations to account for sampling 4 days per week and eligible patients who were not enrolled, by age <1 year (<12 months), 1 year (12–23 months), and 2 to <5 years (24–59 months), with sites as sampling strata. Rates were calculated overall and by viral cause and demographic subsets, with 95% confidence intervals calculated by bootstrap methods based on 1000 bootstrap samples for each rate.16 Demographic groups and high-risk versus low-risk health groups were compared by calculating rate ratios and their bootstrapped 95% confidence intervals.16 Because our study did not directly determine the proportion of children in the county populations with underlying high-risk medical conditions, we assumed that 10% of children who were younger than 5 years were at high risk in both counties on the basis of national data17 and local Rochester data.18
Hospital-Based Percentages of Virus-Positive Patients
The weighted and unweighted frequency distributions of specific virus-associated hospitalizations were comparable and are presented unweighted. The number of virus-specific hospitalizations is the numerator, and the number of enrolled cases is the denominator.
χ2 tests were used to assess differences between enrolled and nonenrolled eligible ARI admissions, associations between demographic and other descriptive characteristics of the enrolled admissions (eg, race/ethnicity, insurance status), and clinical differences between viral causes. Race/ethnicity was categorized as non-Hispanic white, non-Hispanic black, Hispanic (any race), or other.
Before the study was conducted, analysis of hospital databases found that admission rates for ARI were similar for study sampling and nonsampling days. Unpublished analysis of hospitalization data (Statewide Planning and Research Cooperative System hospitalization database19 for New York and a survey of all hospitals in Tennessee) demonstrated that >95% of targeted children who were younger than 5 years and had discharge diagnoses for ARI were hospitalized in the surveillance hospitals in 1999. Hospital administrative databases were audited each month for missed fever or ARI admissions on sampling days.
In 2000, 84790 children who were younger than 5 years resided in the surveillance areas: 46977 children in Monroe County, New York, and 37813 children in Davidson County, Tennessee.20 Although the age and gender distributions for the 2 counties were similar (Table 1) and representative of the entire US population, significant race/ethnicity differences were seen. Davidson County had a higher percentage of non-Hispanic blacks than Monroe County or the United States (32% vs 19% vs 14%), and Monroe and Davidson counties had a lower percentage of Hispanics than the United States (10% vs 8% vs 19%). Of 812 eligible hospitalizations during the 4 weekdays of surveillance between October 1, 2000, and September 30, 2001, 592 (73%) children were enrolled. Of the 220 children not enrolled, 127 (58%) were because of parent/guardian refusals; 44 (20%) were because of parent/guardian unavailability; 40 (18%) were discharged before enrollment; and the remaining 9 (4%) were because of missed admissions, interpreter unavailability, or physician refusal. Twenty children had multiple hospitalizations and accounted for a total of 43 hospitalizations. Enrolled children were comparable to nonenrolled children for gender, race/ethnicity, and county of residence. However, enrolled children were significantly younger and less likely to have private insurance at both sites (data not shown).
Demographic, Clinical, and Risk Factor Assessment
Nearly one third of enrolled children had 1 or more chronic medical conditions. Asthma was an underlying condition for 24% of ARI admissions (10% of those <1 year and 50% of older children). Other medical conditions included chronic pulmonary disease (4%), heart disease (3%), sickle cell disease (2%), immune disorders (1%), kidney disease (1%), cancer (0.3%), and diabetes (0.2%). Sixteen percent of enrolled children were born >1 month early by parental report. Exposures that increased the risk for infection or hospitalization as a result of respiratory viruses were common. Forty-five percent of enrolled children had a smoker in the home, 30% of enrolled children attended child care or preschool >4 hours per week, 56% had another child younger than 5 years in the household, and 42% had another child older than 5 years living at home. Of children who were ≥6 months of age by October 2000 and had a high-risk medical condition for which influenza vaccination is specifically recommended, 36% in Monroe County and only 14% in Davidson County received vaccine according to parental report (P = .005).
Rates of Hospitalization Associated With RSV, Influenza, and Parainfluenza Viruses
One third of enrolled children had 1 of the 3 targeted viral pathogens: 116 (20%) were positive for RSV, 20 (3%) were positive for influenza, and 40 (7%) were positive for PIV. Other viruses were identified in 36% (25 adenoviruses, 19 human metapneumoviruses, and 178 picornaviruses), and no viruses were detected in 39%. Several specimens were positive for >1 virus, but no enrolled child had multiple hospitalizations in which RSV, influenza virus, or PIV were detected. RSV- and influenza-associated hospitalizations overlapped, both peaking in January and February 2001. Influenza isolates were type A (H1N1) and type B, the predominant influenza types circulating in the United States.21 PIV hospitalizations were distributed throughout the year, consistent with other reports22 (Fig 1).
Virus-positive hospitalization rates per 1000 children were 3.5 for RSV, 1.2 for PIV, and 0.6 for influenza virus (Table 2). Rates were not significantly different in the 2 counties (Fig 2), although overall ARI hospitalization rates in Davidson County were 50% higher than those in Monroe County as a result of higher rates of other viruses and virus-negative samples.
Infants who were younger than 6 months accounted for more than half of the RSV hospitalizations and ∼40% of other hospitalizations. Of the targeted viral pathogens identified, 94% of RSV, 80% of influenza, and 85% of PIV were detected in children younger than 2 years (Fig 3). Rates of influenza- and PIV-associated hospitalizations were higher for non-Hispanic black and Hispanic children. In contrast, RSV-related hospitalizations did not differ by race or ethnicity. Non-Hispanic white children were significantly more likely to have private health insurance (60% vs 20%; P < .001), to have a sibling (81% vs 73%; P = .02), and to have a sibling older than 5 years (50% vs 33%; P < .001). Non-Hispanic white children were also less likely to have a high-risk condition (27% vs 36%; P = .03). Children with chronic underlying medical conditions were severalfold more likely to be hospitalized for 1 of the targeted viruses (Table 3), and overall hospitalization rates were higher for boys than for girls.
Clinical Findings in Children With RSV, Influenza Virus, and PIV
The most common admission diagnoses included fever, rule-out sepsis, or febrile seizure (39%); pneumonia (17%); asthma (17%); and bronchiolitis (16%). A significantly higher percentage of influenza-confirmed cases had an admission diagnoses of fever, rule-out sepsis, or febrile seizure compared with other viruses (65% vs 38%; P = .01). A significantly higher percentage of RSV-confirmed cases required supplemental oxygen compared with PIV or influenza cases (61% vs 27% and 30%, respectively; P < .001). The median length of stay for RSV-, PIV-, or influenza associated-hospitalizations was 2 days and did not differ among the viruses. Thirty-one children were admitted to the intensive care unit (5 RSV, 1 PIV, 0 influenza)
On the basis of discharge diagnoses, of the RSV cases, 79% had bronchiolitis and 27% had pneumonia, which were significantly higher than for other causes (P < .05). By comparison, 25% of PIV cases had bronchiolitis and 10% had pneumonia; 20% of influenza cases had bronchiolitis and 5% pneumonia.
This report presents the initial year of comprehensive, prospective, population-based surveillance in children who were younger than 5 years and hospitalized with fever or ARI in 2 geographic regions in the United States conducted through the recently established CDC initiative, the NVSN. The NVSN determined the burden of ARI hospitalization with laboratory-confirmed (viral culture and PCR) RSV, influenza virus, or PIV, all potentially vaccine-preventable diseases.
RSV was the leading viral pathogen associated with ARI hospitalizations in our study and accounted for an estimated 13, 3, and 0.4 RSV-associated hospitalizations per 1000 children younger than 1 year, 1 year, and 2 to <5 years, respectively. Earlier published studies from Medicaid and national databases have estimated RSV-associated hospitalization rates to be between 15 and 40 per 1000 children younger than 1 year, 2 to 10 per 1000 children 1 year of age, and 1 to 2 per 1000 children between 2 and <5 years of age. Although our rates are slightly lower than those previously reported, they are within the same order of magnitude and would be expected to vary somewhat over subsequent surveillance periods with different disease severity. Children who were hospitalized with RSV illness in our study were young; more than half were younger than 6 months, and nearly 80% were younger than 1 year. Many children with RSV received a diagnosis of lower respiratory illness, and more than half required oxygen supplementation. Similar patterns of illness have been reported in earlier studies, underscoring the potential for maternal or neonatal vaccination to prevent RSV disease.10
The influenza-associated hospitalization rates were 1.7, 0.5, and 0.2 per 1000 children younger than 1 year, 1 year, and 2 to <5 years of age, respectively. National and local influenza virus disease reports indicated that the 2000–2001 season was particularly mild, with lower influenza activity than the 3 preceding seasons.21 Two earlier published studies estimated influenza-associated hospitalization rates of 7 to 19 per 1000 children younger than 1 year and 2 to 8 per 1000 children 1 to 2 years of age.2,3 Another report estimated 1.4 to 1.9 influenza-associated hospitalizations per 1000 healthy children younger than 2 years and 0 to 0.3 cases per 1000 healthy children 2 years of age, very comparable to those seen in our study.4 Eighty-percent of the influenza-associated hospitalizations in our study occurred in children younger than 2 years,7 the group recently targeted for influenza vaccination by the Advisory Committee on Immunization Practices and the American Academy of Pediatrics. Similar to other reports, distressingly low influenza immunization rates in the high-risk population groups were noted in both surveillance counties, with significantly lower rates in Davidson County.23–25 Precise estimates of hospitalization rates during both mild and more severe influenza seasons are needed to develop and refine further vaccine policy and will be an important feature of ongoing surveillance efforts in the NVSN.
Estimated PIV-associated hospitalization rates in the NVSN study were 3.2, 1.5, and 0.4 per 1000 children younger than 1 year, 1 year, and 2 to <5 years of age, respectively. An earlier database analysis concluded that PIV 1 to 3 accounted for between 2 and 12 hospitalizations per 1000 US children younger than 1 year and 0.5 to 2 per 1000 children between 1 and 4 years of age, rates similar to ours.6 Thus, it seems that the NVSN documented hospitalization rates comparable to those estimated by database analyses.
Despite the strength of our surveillance network, our studies must be interpreted with several caveats. First, the populations studied may have introduced some systematic biases. Surveillance was conducted in only 2 counties. Although the counties are located in different regions of the United States (Northeast and South) and varied in hospitalization rates and other characteristics, more sites and geographic diversity would add to the generalizability of the data. In addition, ∼5% of children who resided in the 2 surveillance counties may have been excluded from surveillance because they received inpatient care at hospitals in outlying counties, based on databases and surveys conducted in the surveillance areas. Next, enrolled patients seemed to be representative of the total eligible population, but it is possible that the nonenrolled subjects and partial-week surveillance might have introduced bias. Finally, clinical enrollment criteria were based primarily on admission diagnoses recorded by admitting physicians and analyses of discharge diagnoses based on hospital-coded diagnoses. Neither was independently evaluated by study personnel for accuracy or consistency between sites.
Another caveat is that only a single year of surveillance was analyzed. The low influenza hospitalization rates in our study were consistent with national and local reports of lower-than-usual influenza activity. The predominant influenza virus circulating locally and nationally was A (H1N1) with co-circulation of type B. Historically, the greatest numbers of influenza-associated hospitalizations have occurred during epidemics caused by type A (H3N2) viruses.7
Finally, despite the increased sensitivity of PCR methods coupled with culture, some viral agents may have been missed because of low viral titers, instability of RNA, or that some viruses that cause hospitalization have yet to be discovered. Serologic testing might have improved diagnostic sensitivity, but the inclusion of venipuncture would likely have reduced subject participation, and presence of maternal antibodies and asymptomatic seroconversion confounds diagnosis. Specimen collection using nasal and throat swabs may have led to poorer recovery of RSV than using nasal washes or possibly nasopharyngeal swabs. However, as discussed in another report,12 several studies found few differences in viral detection rates among these methods, and nasal swabs are more acceptable to families and less expensive to perform.
In conclusion, this population-based surveillance study revealed a high burden of hospitalization for fever or ARIs associated with RSV, influenza virus, and PIV among young children. Young age, male gender, underlying medical conditions, and black and Hispanic race/ethnicity were significant risk factors for hospitalization. The newly established NVSN, described in this report, has the potential to provide an accurate assessment of the annual variability in viral respiratory disease and an infrastructure to study the impact of currently available and newly licensed viral respiratory vaccines on disease incidence.
This study was supported by the CDC National Immunization Program through cooperative agreements with the University of Rochester (U38/CCU217969) and Vanderbilt University (U38/CCU417958) and by a grant from the National Vaccine Program Office.
NVSN ARI inpatient study collaborators are as follows. University of Rochester: Linda Anderson, Richard Barth, Gerry Lofthus, Anne Mowrer, Kenneth Schnabel, and Andrea Marino; Vanderbilt University: Kathy Holland, Diane Kent, Ayesha Khan, Yi-Wei Tang, Nayleen Whitehead, Sandra Yoder, and Yuwei Zhu; and CDC: Barbara Anderson, Carolyn Bridges, John Copeland, Paul Gangarosa, James A. Mullins, David Shay, and Philip Smith.
- Received September 22, 2003.
- Accepted January 26, 2004.
- Reprint requests to (M.K.I.) Centers for Disease Control and Prevention, National Immunization Program, MS E-61, 1600 Clifton Rd NE, Atlanta, GA 30333. E-mail:
- ↵Erdman DD, Weinberg GA, Edwards KM, et al. GeneScan RT-PCR assay for detection of 6 common respiratory viruses in young children hospitalized with acute respiratory illness. J Clin Microbiol.2003;41 :4298– 4303
- ↵Weinberg GA, Erdman D, Edwards K, et al. Superiority of reverse transcription-PCR to conventional viral culture in the diagnosis of acute respiratory infections in children. J Infect Dis.2004;189 :706– 710
- ↵Mullins J, Erdman D, Weinberg G, et al. Human metapneumovirus infection in children hospitalized for respiratory disease. In: Abstracts of the 40th Annual Meeting of the Infectious Diseases Society of America. Alexandria, VA: Infectious Diseases Society of America; 2002 (abstr 774):180
- ↵The SAS System for Windows [computer program]. Version 8. Cary, NC: SAS Institute Inc; 2002
- ↵Rust KF, Rao JNK. Variance estimation for complex surveys using replication techniques. Stat Methods Med Res.1996;5 :283– 310
- ↵Szilagyi PG, Iwane MK, Schaffer S, et al. The potential burden of universal influenza vaccination of young children on visits to primary care practices. Pediatrics.2003;112 :821– 828
- ↵Statewide Planning and Research Cooperative System (SPARC), New York. Available at: www.health.state.ny.us/nysdoh/sparcs/sparcs.htm. Accessed May 5,2003
- ↵United States Census Bureau. US Census 2000. Available at: factfinder.census.gov/servlet. Accessed May 5,2003
- ↵Centers for Disease Control and Prevention. Influenza activity—United States, 2000–01 season. MMWR Morb Mortal Wkly Rep.2001;50 :207– 209
- ↵Poehling KA, Speroff T, Dittus RS, Griffin MR, Hickson GB, Edwards KM. Predictors of influenza virus vaccination status in hospitalized children. Pediatrics.2001;108(6) . Available at: pediatrics.org/cgi/content/full/108/6/e99
- Copyright © 2004 by the American Academy of Pediatrics