Study design
This study was approved by the Kaiser Permanente Medical Care Program (KPMCP) Institutional Review Board for the Protection of Human Subjects, which has jurisdiction over all the hospitals and clinics described in this study. We conducted a retrospective cohort study that included infants ≥32 weeks GA born and discharged alive from one of 6 Northern California KPMCP hospitals between 1996 and 2004. Infants <32 weeks GA at birth represent a large portion of all infants who receive RSV prophylaxis in our integrated healthcare system and represent only a small portion of all preterm infants born annually [18]. Consequently, infants <32 weeks GA were not included in this study due to routine administration of RSV immunoprophylaxis. To reduce the effects of loss to follow up on our analysis we identified those infants who remained KPMCP members until they were two years of age. We included subjects in the cohort if they were members for ≥20 of the 24 months after the infant’s birth, and if they had no more than two consecutive months without membership. Subjects were defined as members for each month in the study if they were either insured by KPMCP or used KPMCP medical services. See Figure 1 for flowchart of the cohort selection.
We extracted data regarding these infants from KPMCP databases, including inpatient and outpatient encounters and diagnoses. Because there may be some use of KPMCP facilities without membership, we conducted a sensitivity analysis examining our results among a subcohort of infants who were recorded as members for all 24 months after the infant’s birth (see Additional file 1: Tables S1, S 2, and S 3). The race/ethnicity of participants was based on the self-reported race/ethnicity of infants' mothers.
Variables and measurements
To identify only episodes that included a true diagnosis of bronchiolitis while capturing all relevant clinical information associated with a bronchiolitis episode, we classified relevant International Classification of Diseases (ICD-9) codes into two categories. Category A included diagnostic codes specific for bronchiolitis in children younger than two years (ie, 466.1 acute bronchiolitis, 466.1x acute bronchiolitis organism specified, 466.0 acute bronchitis, 480.0–480.2 pneumonia due to adenovirus, RSV or parainfluenza, 079.0 adenovirus and 079.6 RSV). Category B included diagnostic codes likely to be highly related to bronchiolitis in a child with a current diagnosis of bronchiolitis (i.e., 786.06 tachypnea, 786.07 wheezing, 786.09 other respiratory distress, 780.6–780.60, 786.2 cough, 460 acute nasopharyngitis, 465 acute upper respiratory infections, 465.9 acute upper respiratory infection not otherwise specified, and 487.1 acute upper respiratory infection due to influenza.)
We defined a bronchiolitis episode as a period that included a diagnostic code specific for bronchiolitis from Category A, began either with the first bronchiolitis-related diagnostic code from Category B occurring within 2 days prior to the first Category A code or with the first Category A code if there was no Category B code within 2 days prior to the first Category A code, and ended with a diagnostic code from either Category A or Category B followed by 14 clear days without a bronchiolitis or related diagnosis. An episode of care can include multiple healthcare visits and/or various types of healthcare visits, including hospitalization or outpatient visits with the aforementioned category codes. The duration of an episode of care is defined as the length of time between the first and last visit.
For all infants, we extracted data regarding additional clinical predictors including GA, sex, race/ethnicity, maternal age, birth weight, a diagnosis of BPD or congenital anomaly (ICD-9 codes 425.3x, 425.4x, 425.8x, 745.xx, 746.xx, 747.xx) and the number of siblings present in the home (≥1 sibling <5 years of age). For hospitalization records, we also determined whether the infant required assisted ventilation (ICD-9 codes 93.90, 93.91, and 96.7x).
We classified infants as being small for gestational age (SGA) (<5th percentile) using birth weight and GA according to the algorithm developed by Brenner et al. [19]. We fitted a smooth cubic spline as a function of total oxygen exposure separately for births occurring at 32–37 and ≥37 weeks GA. Both curves showed a flat effect for total oxygen exposure <200 hours and a clear increase at >200 hours, suggesting a step function and no interaction between GA and total length of oxygen therapy. For infants treated in the neonatal intensive care unit, we created an oxygen exposure variable: no supplemental oxygen exposure during the neonatal period, no bronchopulmonary dysplasia (BPD); supplemental oxygen exposure of 1 to <200 hours during the neonatal period, no BPD; supplemental exposure ≥200 hours during the neonatal period, no BPD; and BPD. We classified infants who did not require intensive care during the neonatal period as having no supplemental oxygen exposure.
Parental asthma history was established by scanning the parent’s records for the period 18 months before to 6 months after the infant’s birth and determining whether the parent had ≥2 clinical visits 14 days apart with an ICD-9 code (493.xx) for asthma, and/or the parent’s electronic record listed asthma on their Significant Problem List.
Statistical analysis
All statistical analyses used SAS (version 9.1, SAS Institute, Inc., Cary, NC) or Stata (version 9.2, Stata Corp., College Station, TX). After identifying duration of episode, we eliminated episodes lasting >60 days from further analysis because it was unlikely that these represented isolated bronchiolitis (see Additional file 1: Table S4). Manual chart review of infants with episodes lasting >60 days confirmed that these episodes represented hospitalizations for which bronchiolitis was not the primary diagnosis. Infants remaining after this step constituted the final analysis cohort. We used descriptive statistics to report frequency and duration of bronchiolitis episodes and frequency of hospitalization associated with the episode, by age at the episode, gender, GA, and race/ethnicity. We used a moving average of GA and GA minus 1 week to develop a graph of mean, 10th percentile and 90th percentile for the duration of EOC for each GA. We used multivariable logistic regression to assess the effect of GA, gender, race/ethnicity, family history of asthma, neonatal oxygen use and other clinical predictors on the outcome of whether an infant had a documented bronchiolitis EOC and on the outcome of whether an infant had a documented bronchiolitis EOC lasting >1 day. We used White race/ethnicity as the referent category for this multivariate analysis because this was the largest population in our cohort and because previous literature has described disparate risk of bronchiolitis for non-White racial/ethnic groups. We calculated the relative contribution of each predictor using the differences between the log likelihood of the full model and the log likelihood of a model without each of the predictors, and the relative contribution of each predictor was defined as the ratio of its log likelihood difference to the sum of the likelihood differences from all predictors × 100 [20].