Prevalence of infant bronchiolitis‐coded healthcare encounters attributable to RSV

Abstract Aim We sought to determine the proportion of bronchiolitis episodes attributable to respiratory syncytial virus (RSV) among ICD‐9 coded infant bronchiolitis episodes which were tested for RSV. Methods Bronchiolitis healthcare encounters were extracted from Kaiser Permanente Northern California databases for years 2006 to 2009. We used ICD‐9 codes for bronchiolitis to capture bronchiolitis‐related healthcare encounters including hospital admissions (Hospitalization), emergency department visits (EDV), and outpatient visits (OPV). We reported the monthly proportion of RSV‐positive bronchiolitis episodes among tested bronchiolitis episodes. We used logistic regression to assess association between bronchiolitis episodes and patient demographic and health care characteristics. We also used logistic regression to assess association between decision to test and patient demographics and health care characteristics. Results Among 10,411 ICD‐9 coded infant bronchiolitis episodes, 29% were RSV tested. Fifty one percent of those tested were RSV positive. Between December and February, and in infants ≤6 months, the proportion of bronchiolitis episodes that were attributable to RSV was 77.2% among hospitalized episodes, 78.3% among EDV episodes, and 60.9% among OPV episodes, respectively. The proportion of RSV‐positive bronchiolitis episodes varied based upon infant age at diagnosis, level of health care service used, and time of the year of the episode. Conclusion Estimation of the proportion of ICD‐9 coded bronchiolitis episodes attributable to RSV is more specific when restricting to bronchiolitis episodes during peak months, younger infant age, and those requiring higher level of healthcare.

diseases and health conditions across all claims. ICD-9 codes were in use before converting to ICD-10 version in 2015; while the codes for bronchiolitis have changed, their descriptors have not.
Encounters that occurred within 14 days of each other were considered as one episode and were classified based on the highest level of care visit type the infant experienced (Hospitalization > EDV > OPV), while the bronchiolitis encounters that were 14 days or more apart were treated as independent episodes. More than one test and/or one detection method (antigen, culture, and Polymerase Chain Reaction [PCR]) could be performed during one bronchiolitis episode. The episode was classified as RSV positive if any of the tests were positive. If multiple tests per episode were negative, only the PCR result was retained. Since 2006, KPNC started using primarily PCR to test for RSV (78.3%), with antigen and culture methods constituting only 13.2% and 8.5% of the tests, respectively. In this study, we included ICD-9 10 -identified infant bronchiolitis episodes captured in the PRIMA cohort from 2006 to 2009 and which were also tested for RSV.
We reported the monthly proportion of RSV-positive bronchiolitis episodes among tested bronchiolitis episodes. The proportion of RSVpositive bronchiolitis episodes was further calculated among the subset of tested bronchiolitis episodes: by infant age at the time of bronchiolitis diagnosis (≤6 months and ≤ 3 months), by both infant age at the time of bronchiolitis diagnosis (≤6 months) and the severity of the bronchiolitis episodes (OPV, EDV, and Hospitalization), and by gestational age (term vs. preterm [gestational age < 37 weeks]). We conducted post hoc analysis using multivariable logistic regression to assess the association of infant age at bronchiolitis diagnosis, sex, season of the episode (peak RSV season vs. non-peak season), severity of the bronchiolitis episodes (OPV, EDV, and Hospitalization), RSVimmunoprophylaxis eligibility based on American Academy of Pediatrics (AAP) guidelines, 11,12 and RSV-immunoprophylaxis receipt (ever receipt and never receipt) with outcome of RSV positivity. Lastly, we used multivariable logistic regression to identify factors that were associated with testing for RSV. Analyses were conducted using Stata software version 15 (College Station, TX). 13
In total, 29.2% (n = 3,040 episodes among 2,057 infants) of all ICD-9 identified bronchiolitis episodes during infancy were laboratory tested for RSV. The proportion of bronchiolitis episodes that were tested within each of the three ICD-9 codes was as follows: among 608 RSV-bronchiolitis coded episodes, 59.4% were tested; among 25 RSV-pneumonia coded episodes, 44.0% were tested; and among 9,778 non-RSV coded bronchiolitis episodes, 27.3% were tested (   (Table S2). Further analysis limiting the bronchiolitis episodes to those episodes which occurred between December and February and in infants ≤6 months, the proportions of RSV positive bronchiolitis episodes were 77.2% among hospitalized episodes, 78.3% among EDV episodes, and 60.9% among OPV episodes, respectively ( Table 2).
We next stratified RSV tested bronchiolitis episodes by established risk factors for RSV. A stratified analysis was conducted on bronchiolitis episodes occurring during the first 6-months by gestational age, a known risk factor for RSV infection 14,15 and an important factor determining eligibility and receipt of RSV-immunoprophylaxis. 16  We conducted a multivariable logistic regression for outcome of RSV positive bronchiolitis episodes among tested episodes (Table 3).
Younger age (in months) at the episode (adjusted odd ratio    (please see Table S2 for full year results).

| DISCUSSION
In a large population-based study of infants who were continuously enrolled in an integrated healthcare system, we calculated the proportion of ICD-9 defined and laboratory tested bronchiolitis episodes that were attributable to RSV. ICD-9 coded bronchiolitis episodes occurring during the peak months of winter viral season, among younger infants, and among those with higher levels of care, were more likely to be attributable to RSV.
In agreement with other studies, 7,17 the majority of ICD-9 coded and RSV tested bronchiolitis episodes during infancy were associated with RSV infection (51.4%). Our study confirms that the proportion of RSV positive bronchiolitis episodes is higher (73.6% of RSV tested bronchiolitis episodes) during the peak months of winter viral season (December to February), among infants under 6 months of age, and among those hospitalized for bronchiolitis. However, the proportion of bronchiolitis episodes that were tested for RSV was also higher between the months of December and February. The monthly ratio of the proportion of RSV positive bronchiolitis episodes to the proportion of RSV ever tested bronchiolitis episodes exceeded one between November and March, and was less than or close to one the remainder of the year (Figure 2). However, testing in our study is biased by well recognized RSV risk factors or clinical presentations, such as younger age, more severe clinical disease, and prematurity, among others factors. Many of these factors are also more likely to be associated with RSV than other viral etiologies of respiratory illness in infancy. These pretest features inform clinical testing and, thus, pretest probability of having RSV, which increases the specificity of use of electronic record coding for RSV bronchiolitis for purposes of RSV surveillance.
In a multi-year surveillance study of 31 US ED sites, Makari and colleagues reported that 62% of ICD coded bronchiolitis EDV encounters were RSV positive by PCR during the peak period (January 15 to end of February). 18 Hall and colleagues also reported in a multi-year prospective study that 26% of acute respiratory illnesses in children age < 24 months and hospitalized during October through March were attributable to RSV. 19 The fact that the proportion of RSV positive bronchiolitis episodes in EDV and hospitalization were similar in our analysis shows that results from a prospective study by Makari  The majority of the infants in our study were deemed to be low-risk (97%), born at term or later, and without underlying conditions. Three percent of episodes were among those born preterm, and RSVimmunoprophylaxis eligible (high-risk) infants who may or may not have received RSV-immunoprophylaxis. In our multivariable regression analysis, there were decreased relative odds of having at least one RSV positive bronchiolitis episode for RSV immunoprophylaxis eligible vs. noneligible and for subjects receiving RSV-immunoprophylaxis. Eighteen percent of those eligible to receive RSV immunoprophylaxis did not receive any RSV-immunoprophylaxis and 1% of those ineligible, received at least one dose of immunoprophylaxis. It may be that the routine care of high-risk infants is different from the care for term or later birth infants. Parents may be less likely to send high-risk infants to daycare, may take extra caution with regard to ill contacts, and healthcare providers may do more testing to include a larger proportion of infants with a broader spectrum of respiratory illness, as evidenced by the higher proportion tested (34.6% vs. 28.3%).
Our study has several strengths, including a large sample of bronchiolitis episodes which were tested for RSV. In addition, the majority of tests (78.3%) were done using the most sensitive clinical testing modality for RSV detection, PCR. Antigen and culture RSV detection results were used only when PCR was not performed. This approach

| CONCLUSION
Estimates of RSV-bronchiolitis burden and seasonal distribution are often based on ICD-9 codes for bronchiolitis. We provide estimates of ICD-9 codes for bronchiolitis being attributable to RSV based on results confirmed using RSV laboratory testing. The proportion of RSV positive bronchiolitis episodes varies by infant age at the illness, level of health care service used, and time of the year of the episode.
Estimation of RSV positive bronchiolitis based on ICD-9 codes can be improved by restricting to episodes occurring during peak months of winter viral season, among infants ≤6-months age at time of illness, and among those requiring higher levels of healthcare. Strategies to improve the specificity of bronchiolitis-coded episodes have implications for public health surveillance, the design of prevention strategies, and for research. As an example, identifying RSV peak season and target age group may help to efficiently channel scarce public health resources, as well as to implement prevention measures such as daycare closures or the testing of targeted prevention measures, to prevent disease spread. In addition, this information may inform the start and end of use of RSV immunoprophylaxis in high risk populations, improving prevention during periods of highest risk of infection, and potentially shortening duration, and, thus, cost of RSV immunoprophylaxis. Lastly, among those tested, the majority of RSV positive bronchiolitis episodes were among infants born at term or later gestational age, reinforcing that the greatest disease burden is among term infants.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE
This study was approved by both the Kaiser Permanente Northern

California (KPNC) and Vanderbilt University Institutional Review
Boards.