Population-based trends and underlying risk factors for infant respiratory syncytial virus and bronchiolitis hospitalizations

Objective Respiratory syncytial virus (RSV) is a common pathogen during infancy, with the potential to cause serious disease and mortality in high-risk groups. The objective of this study was to characterize trends of RSV and bronchiolitis hospitalizations in the first year in a population-based cohort and assess differences in trends according to risk status. Methods Using an observational retrospective cohort design, we examined a California population-based dataset of vital statistics linked to hospital discharge data for up to 1 year after birth from 1997–2011. Infants were categorized by medical condition and then by gestational age. Medical conditions of interest included chronic lung disease, certain congenital heart diseases, or others known to affect risk for developing severe bronchiolitis. The primary outcome was hospitalization due to RSV; secondary outcome was hospitalization for unspecified bronchiolitis (UB) not coded as RSV. Annual person-year rates were calculated for infants <12 months of age during January to December of each year. Results Of 7,298,401 infants born during the study period, 121,230 (1.7%) had a medical condition associated with risk; these infants experienced 6853 RSV and 6568 UB hospitalizations in the first year. In infants without medical conditions, 96,694 RSV and 69,886 UB hospitalizations occurred. All-cause infant hospitalizations declined over time from 12.2 to 9.3 per 100 person-years. RSV hospitalization rates for infants with medical conditions decreased from 7.6 to 3.4 per 100 person-years, with the largest relative decline in infants with chronic lung disease (12.0 to 5.0 per 100 person-years). For infants without medical conditions, RSV hospitalizations declined from 1.4 to 0.8 per 100 person-years, with greater decreases among preterm infants with earlier gestational age. UB hospitalization rates remained relatively stable across the study years, from 6.2 to 5.4 and 1.0 to 0.8 per 100 person-years for infants with and without medical conditions. Conclusions Various interventions may have contributed to observed decreases in RSV hospitalizations from 1998–2011, which were greater in high-risk populations recommended for RSV immunoprophylaxis and not observed with UB. Further efforts to promote evidence-based practice and optimal targeting of appropriate interventions will ensure continued improvement in care for vulnerable infants.


Results
Of 7,298,401 infants born during the study period, 121,230 (1.7%) had a medical condition associated with risk; these infants experienced 6853 RSV and 6568 UB hospitalizations in the first year. In infants without medical conditions, 96,694 RSV and 69,886 UB hospitalizations occurred. All-cause infant hospitalizations declined over time from 12.2 to 9.3 per 100 person-years. RSV hospitalization rates for infants with medical conditions decreased from 7.6 to 3.4 per 100 person-years, with the largest relative decline in infants with chronic lung disease (12.0 to 5.0 per 100 person-years). For infants without medical conditions, RSV hospitalizations declined from 1.4 to 0.8 per 100 person-years, with greater decreases among preterm infants with earlier gestational age. UB hospitalization rates remained relatively stable across the study years, from 6.2 to 5.4 and 1.0 to 0.8 per 100 person-years for infants with and without medical conditions. PLOS

Introduction
Respiratory syncytial virus (RSV) is a leading cause of hospitalization and can result in serious morbidity and mortality in infants, particularly infants born preterm or with a high-risk medical condition such as chronic lung disease (CLD) or congenital heart disease (CHD) [1][2][3][4].
Because of the high burden of disease, interventions to prevent serious RSV disease have been implemented in the past 2 decades for these high-risk infants [5][6][7]. There have been reports of reduced incidence of RSV and/or bronchiolitis hospitalization in recent years for infants overall and specifically for infants with CLD and CHD [1,8,9]. However, none of these studies have employed a cohort design in which the population is prospectively followed forward in time; instead, they have used estimated denominator populations based on national birth statistics. Furthermore, due to this design limitation, no study has examined hospitalization rates among preterm infants because gestational age (GA) at birth was not available for hospitalized infants. It is not clear whether observations of reduced hospitalizations for RSV and bronchiolitis may be due to improvements and changes in the care of neonatal and infant respiratory disease in general, or whether prevention and intervention measures targeted directly at RSV have been particularly effective. Specifically, RSV immunoprophylaxis of high-risk infants with palivizumab was increasingly utilized in the United States during the study period following its licensure in 1998. In this study, we assessed infant hospitalization trends for RSV and unspecified bronchiolitis (UB) by presence of higher-risk medical conditions and GA during the first year of life in a population-based cohort. We compared trends for those infants to those without higher-risk medical conditions. By using statewide California databases linked with patient identifiers, we tracked infants from birth through the first year of life and accounted for admissions to any hospital in the state.

Methods
California hospital discharge data from the Office of Statewide Hospital Planning & Development (OSHPD) linked to vital statistics for 1997-2011 were used to create a longitudinal, population-based cohort of newborns followed to 1 year of age. These years were those for which complete data were available. The OSHPD database of hospital discharges captures almost all hospital births and subsequent hospital admissions in California, with the exception of military hospitals or birthing centers not reporting to OSHPD (<3% of all deliveries). Although there is a separate record for each hospitalization, a unique identifier for each individual allows matching of patients across their first year after delivery. The probabilistic linkage algorithm used to perform the patient match by OSHPD is robust and described in detail previously [10]. In addition to various sociodemographic characteristics, each record contains a list of International Classification of Diseases-9 (ICD-9) codes to characterize the admission. Birth hospitalizations were identified using ICD-9 codes V30.XX-V39.XX as in previous work and included delivery hospitalizations as well as subsequent transferred admissions [1]. California birth hospitalizations regardless of state of residence were included; the large majority (>98%) were California residents. We included infants who had GA of 22-44 weeks recorded on their birth certificates. Individuals with potential errors in coding GA, death during delivery or during the birth hospitalization, or inconsistent data, such as admission occurring after the date of death, were excluded.
Infants were categorized according to medical condition as noted in S1 Table. Any diagnosis occurring at any time point during the first year of life was assigned as a characteristic of that infant. CHD diagnoses were characterized into high-risk or low-risk strata consistent with previous research [1], with high-risk superseding low-risk status. As a category for analysis, Down syndrome without CHD was considered separately and was mutually exclusive of higher-risk or lower-risk CHD. A category of "other high-risk" was created, encompassing cystic fibrosis with pulmonary manifestations, neuromuscular disease, human immunodeficiency virus, immunodeficiency, and congenital and metabolic disease. The other high-risk group excluded infants with CLD, high-risk or low-risk CHD, Down syndrome, and congenital anomalies of the respiratory system.
For the analyses of risk among infants by GA without other serious medical conditions, infants with CLD, congenital anomalies of the respiratory system, higher-risk CHD, or the other high-risk conditions were excluded. Down syndrome and lower-risk CHD were not excluded for this analysis, as these conditions have not traditionally been targeted for RSV-specific prevention efforts. A sensitivity analysis excluding lower-risk CHD and Down syndrome was performed to compare the results in infants without any comorbid conditions.
The outcomes of interest were hospital admission related to either RSV or UB occurring prior to 12 months of age. We identified transfers based on discharge from one hospital and admission to another hospital on the same date; transfers were classified as a single hospital admission. Patients with any relevant ICD-9 codes during the hospitalization for RSV (480.1, 466.11, 079.6) or UB (466.19, 466.1) listed as a diagnostic code were identified, with those having both considered as having RSV.
Annual trends of RSV-and UB-coded nonbirth hospital admissions during the first year after birth were compared by medical condition and then for those without medical conditions by GA category (22-28, 29-31, 32-34, 35-37, and 38-44 completed weeks GA). When presenting results for trends over time with years as the unit, the birth year of the patient was the basis for categorization. For example, if a nonbirth hospitalization for an infant born in November 2011 occurred in January 2012, this would be categorized as a 2011 patient and hospitalization. Infants were followed for 12 months or until death if this occurred within the first year of life. During follow-up, all hospitalizations, including repeat hospitalizations, were considered to be the event. Hospitalization rates were estimated per 100 person-years and were assessed for trends over time using unadjusted Poisson regression. By evaluating person-time incidence, we were able to adjust the exposure denominator for infants who died and were no longer eligible to be hospitalized, as well as adjust the numerator for infants who may have been hospitalized more than once for RSV or UB.
Analyses were conducted using SAS 9.4 software (SAS Institute, Cary, NC). This study was approved by the Stanford University Institutional Review Board. The patient data used in this study were fully anonymized prior to our access, and the Institutional Review Board waived the requirement for informed consent.

Results
After exclusions, 7,298,401 infants were born and discharged to their homes after the birth hospitalization during the study period (Fig 1). In this cohort, 121,230 infants (1.7%) had an identified medical condition (Table 1). In this cohort of infants with medical conditions, 95,785 nonbirth hospitalizations occurred during the first year of life; there were 6853 (7.2%) hospitalizations for RSV and 6568 (6.9%) for UB. The proportion of infants hospitalized for RSV and UB were 5.4% (6580 infants) and 4.9% (5915 infants), respectively in this cohort. Among infants without identified medical conditions, 712,135 nonbirth hospitalizations occurred: 96,694 (13.6%) for RSV and 69,886 (9.8%) for UB. The proportion of infants hospitalized was 1.3% (95,270 infants) for RSV and 0.9% (66,891 infants) for UB. Although RSV and UB were more common reasons for hospitalization in the cohort without medical conditions, hospitalization rates were higher for infants with medical conditions (5.9 per 100 person-years for both RSV and UB) compared with those without (1.3 and 1.0 per 100 person-years for RSV and UB, respectively). All-cause infant hospitalization declined by 24%, from 12.2 per 100 person-years in 1998 to 9.3 per 100 person-years in 2011 (ptrend <0.001).
To assess changes associated with years in which RSV immunoprophylaxis was increasingly used among high-risk populations [11,12] <0.001). The rates of UB hospitalizations among infants with identified medical conditions showed less change (Fig 2C).
For infants without medical conditions, rates of RSV hospitalizations were highest for the most preterm infants and decreased with increasing GA (Fig 3A). RSV hospitalization rates decreased for all preterm GA categories, with greater decreases in those with earlier GA: 22-28 weeks (from 2.  Rates of UB remained more stable across GA groups (Fig 3B). A similar pattern of decreased risk by older GA was seen for UB hospitalizations. The total number of UB hospitalizations was 53,475 for infants 38-44 weeks GA and 16,411 for infants 22-37 weeks GA. RSV represented a lower proportion of all bronchiolitis (RSV or UB) hospitalizations among more preterm infants.
RSV and UB hospitalizations by sociodemographic and clinical characteristics are shown in Table 2. Privately insured patients were less likely to be hospitalized in general, but particularly so for RSV and UB. This was true for infants with and without identified medical conditions.

Discussion
We evaluated trends in admissions for RSV and other bronchiolitis for neonates born in California, stratified by risk status. RSV and UB hospitalizations are common in the first year of life. Consistent with previous reports, RSV-coded hospitalizations have declined over the past decade. This was true for the general population of infants, and the greatest declines were observed for infants with CLD, high-risk CHD, and early preterm birth. On the other hand, admissions for RSV in term infants without medical conditions and for UB across all infant groups remained relatively stable from 1998 to 2011.
Various interventions, including RSV immunoprophylaxis, may have contributed to the observed declines in RSV hospitalizations. In a study of patients with significant heart disease using a US national sample of pediatric hospital admissions, a 36% decrease in RSV hospitalizations was seen when comparing pre-and post-palivizumab guidelines for that population [9]. A similar finding was observed for high-risk CHD and CLD in a second study [1]. Although we did not have patient-level data on RSV immunoprophylaxis, other studies have demonstrated trends of increasing palivizumab use between 1998 and the 2004-2006 timeframe [11,12]. These trends correspond to the temporal trend of declining RSV admissions from 1998 to 2006 seen in our results for infants with medical conditions and early gestational age. We speculate that infants with other high-risk conditions that have not been explicitly targeted for immunoprophylaxis may also have had increased usage of palivizumab due to recognition of their high-risk medical status for respiratory illness. Continuing to target high-risk groups with optimal treatment strategies and implementation of quality improvement should lead to sustained public health benefit.
Breastfeeding is considered a public health benefit in part due to its potential for decreasing infections in infancy [13]. National trends in breastfeeding increased after 1999 and extend beyond the end of our study period [14,15]. Increases in breastfeeding may have contributed to declines in RSV hospitalizations during the study period. Presumably the increase in breastfeeding and the protection against infection that it confers would have been reflected in the      number of UB hospitalizations; however, the decline was not as pronounced during the study period. Furthermore, one study indicated a significantly lower rate of exclusive breastfeeding in preterm infants (8% vs 19%) [16], and high-risk infants may be less likely to sustain breastfeeding due to their medical complexity. We found that infants who were covered by Medicaid had higher rates of RSV and UB hospitalization (Table 2), a finding consistent with other studies [3]. These findings help validate this study's analytic approach of using linked California datasets to evaluate infant hospitalization rates.
Although this and previous studies have shown declines in all infant RSV and UB hospitalizations between 1998 and 2012, the magnitude of the declines have differed across studies [1,8]. In the current study, RSV and UB hospitalizations declined by approximately 38% and 25%, respectively, among infants without medical conditions, in the context of a 25% decline in infant hospitalizations due to any cause. Doucette et al found that RSV and UB hospitalizations decreased by 10% and 34%, respectively, among infants without medical conditions between 1997 and 2012, in the context of a 25% decline in all-cause infant hospitalizations [1]. Hasegawa et al demonstrated a 21% reduction in infant bronchiolitis hospitalizations due to RSV and other causes between 2000 and 2009 [8]. The larger decline in RSV hospitalizations in the current study may be due to decreased testing for RSV during the study interval in California relative to other parts of the country, perhaps due to an increased prevalence of managed care and adoption of recommendations against routine testing for RSV, which would in turn lead to a decrease in RSV-coded respiratory illnesses [17].
Although there was a general decline in RSV hospitalizations during the study period, there was a sudden spike in RSV hospitalization rates in 2010 for most groups (Figs 2C and 3A), with a spike also seen among infants born in 2009 at <28 weeks GA. However, by 2011, hospitalization rates were again consistent with the general trajectory observed in previous years. It is unclear why this spike in 2010 occurred. One speculation is that this was due to the timing of the 2009-2010 influenza pandemic, which likely inhibited RSV circulation and increased concern and confirmatory testing for viral respiratory disease, leading to increased incidence and diagnosis of RSV hospitalizations among at-risk groups the following RSV season. Additional factors such as seasonal differences in RSV strain pathogenicity could have also played a role. As additional years' data become available, it will be important to evaluate if 2010 is an anomaly or part of a trend of increasing annual variability. If the spike was due to increased testing during the 2009-2010 pandemic, the disease rates in 2010 can provide some insight into what RSV and UB hospitalization rates might be if there was more comprehensive laboratory testing for RSV.
A limitation of this study that should be considered is that this was an analysis of California hospitalizations of infants born in the state, and admissions could not be captured when infants moved out of state or had hospitalizations outside of California. It should also be noted that given the sample size, even small differences were expected to be statistically significant; as a result, the primary presentation of the results focuses on clinically meaningful differences rather than on statistical significance.
In summary, we found a general decline in RSV hospitalizations across a large populationbased cohort. This was particularly true for high-risk conditions that are recommended for RSV immunoprophylaxis, such as CLD, high-risk CHD, and preterm birth at early GA. During this study period, the level of hospitalizations for UB remained relatively stable. As there appear to be interventions and preventive efforts that may alter outcomes, it will be important to continue research into which populations will benefit from targeted interventions and implement appropriate measures to provide maximal public health benefit.
Supporting information S1 Table. Medical conditions for categorization. a Infants with these conditions were excluded from analyses of risk by gestational age among infants without serious medical conditions. (DOCX)