Variation in Inhaled Medication Use for Infants With Congenital Heart Disease and Bronchiolitis Free

We conducted a multicenter, retrospective cohort study using data from the Pediatric Health Information System (PHIS) database from January 1, 2015 to June 30, 2019. PHIS is an administrative database with resource utilization data from 52 participating tertiary children’s hospitals. Participating sites submit patient demographics, discharge diagnosis and procedure codes, and detailed billing data from inpatient, observation, emergency department, and ambulatory surgical encounters. Data are deidentified at the time of entry into the database and are subjected to rigorous quality checks before inclusion. For our study, we used data from 51 out of the 52 hospitals that had all relevant data available. IBM Watson Health (Ann Arbor, MI) manages the data warehouse function, and the database is operated by the Children’s Hospital Association (Lenexa, KS).

Our sample consisted of infants ≤12 months old with congenital heart disease hospitalized with bronchiolitis. We defined a bronchiolitis hospitalization as a hospitalization with International Classification of Diseases (ICD) codes consistent with either (1) a primary diagnosis of viral bronchiolitis (466.11, 466.19, J21.0-J21.1, J21.8- J21.9) or (2) a primary diagnosis of a symptom of bronchiolitis (tachypnea, dyspnea, cough, acute respiratory failure, or fever) and a secondary diagnosis of bronchiolitis. We defined CHD as having a secondary ICD, Ninth or Tenth Revision, Clinical Modification (ICD-9-CM or ICD-10-CM) code consistent with CHD (745.0- 747.49, Q20-Q26.9). We excluded infants who were transferred from inpatient to inpatient because we could not determine treatments received before arrival at a PHIS-participating hospital. For infants with >1 hospitalization, only their index hospitalization was included in the sample as medication usage patterns for an initial admission may differ from those for a readmission.

We conducted an initial PHIS query to assess the overall use of specific inhaled medications: racemic epinephrine, hypertonic saline, and β-2-agonists in this population. Hypertonic saline and β-2 agonists were the most used interventions, so we elected to specifically assess the use of these further in this study. The use of racemic epinephrine in our sample was lower (∼11%) with less hospital variation (2% to 29% of cases), so we assessed the use of hypertonic saline and β-2 agonists in this study.

The primary exposure was the proportion of days that patients received inhaled β-2-agonists (albuterol or levalbuterol) or hypertonic saline at each hospital (ie, hospital-level proportion of days) as opposed to medication usage for each individual patient (ie, patient-level proportion of days). For example, if a hospital cared for eligible infants for 1000 days over the study period, and these infants received a β-2-agonist on 250 of those days and hypertonic saline on 300 of those days, a hospital-level exposure of 25% was calculated for β-2-agonists and 30% for hypertonic saline. Demographic covariates included age in months, sex, race (non-Hispanic White, Non-Hispanic Black, Hispanic, other/unknown), payer (government, commercial, other), median household income in quartiles, admission quarter, admission year, and admission source (transfer other than inpatient to inpatient, direct admit, unknown). Clinical covariates included ICD-9/ICD-10 codes (see Supplemental Table 4) representing severe or complicated CHD, including the presence of critical CHD (conditions that require surgical or cardiac catheterization intervention within the first year of life⁷ ), the presence of single ventricle CHD, the presence of congestive heart failure, the presence of pulmonary hypertension, complex chronic conditions as defined by Feudtner et al⁸  (using ICD-9/ICD-10 codes to identify medical conditions of expected 12 months duration that likely require specialty pediatric care or tertiary care hospitalization), the need for extracorporeal membrane oxygenation (ECMO), and severity of illness defined by Hospitalization Resource Intensity Score for Kids⁹  (H-RISK; calculated using cost and All Patient Refined Diagnosis Related Group Severity of Illness score). H-RISK scores compare the relative intensity of hospital care for inpatient pediatric populations, with a higher score representing a higher intensity of care. Scores can range from as low as 0.18 (normal newborn), to as high as 91.66 (heart and/or lung transplant). For context, the H-RISK score for an uncomplicated bronchiolitis admission (APR DRG 138) with minor severity of illness is 0.49.⁹ 

Outcomes included length of stay, 7-day all-cause readmission, ICU use, and mechanical ventilation use at each hospital (ie, hospital-level outcomes) using existing flags for these variables as defined in the PHIS database. We did not attempt to measure noninvasive ventilation use because there is no reliable way to capture this with administrative data.

Patient-level demographics and clinical characteristics of the study cohort were summarized by using frequencies and percentages for categorical variables and medians with interquartile range (IQR) for continuous variables. We assessed hospital-level variation in the proportion of days on β-2-agonist and/or hypertonic saline using logistic regression models. We adjusted hospital-level β-2-agonist use, hypertonic saline use, 7-day all-cause readmission, ICU use, and mechanical ventilation use for differences in patient and clinical characteristics across hospitals using generalized linear mixed models assuming an underlying binomial distribution. We adjusted hospital-level length of stay for differing patient and clinical characteristics using log-linear models. All adjustment models included a random effect for patients clustered within a hospital. After assessing adjusted hospital-level outcomes for normality, we used linear regression models to assess (1) the hospital-level association of days on β-2-agonist with days on hypertonic saline and (2) the hospital-level association of days on β-2-agonist and days on hypertonic saline with LOS, 7-day all-cause readmission, mechanical ventilation use, and ICU use.

All analyses were performed by using SAS v 9.4 (SAS Institute, Cary, NC). Two-sided P values <.05 were considered statistically significant. This project was deemed exempt by the Children’s Hospital Los Angeles Institutional Review Board (per 45 CFR 46.102(f)) and was approved by the leadership of the Children’s Hospital Association and the Pediatric Research in Inpatient Settings Executive Council.

We identified 6846 index hospitalizations for bronchiolitis in infants with CHD from 52 hospitals, with a median (IQR) of 116 (64 to 177) patients admitted at each hospital. The median age was 4 months, and the majority were male (55.8%; n = 3821), Non-Hispanic White (44.5%; n = 3044), with government insurance (68.7%; n = 4701; Table 1). There was no difference noted in the frequency of CHD and bronchiolitis diagnoses noted in our samples before and after the transition from ICD-9-CM to ICD-10-CM diagnoses codes.

Some patients had markers of more severe CHD, including critical CHD (12.6%; n = 864), single-ventricle disease (3.1%; n = 215), and congestive heart failure (8.6%; n = 591; Table 2). Utilizing the definition of complex chronic condition by Feudtner et al,⁸  almost one-half (49.0%; n = 3357) had a cardiovascular complex chronic condition, which includes both complex congenital and noncongenital heart diseases. Overall, 21.6% (n = 1476) had other congenital complex chronic conditions, whereas 10% (n = 687) had respiratory complex chronic conditions.

Nearly one-half (2974 infants, 43.4%) of our sample received a β-2-agonist during hospitalization, and 1554 (22.7%) received hypertonic saline. In our unadjusted model, the hospital-level proportion of days of β-2-agonist and hypertonic saline use varied across hospitals from 1.3% to 69.6% (P <.001) and 0.0% to 78.4% (P <.001), respectively, with a positive association noted between the proportion of days of hypertonic saline and β-2 agonist use (r² = 0.17, P = .003). In our covariate-adjusted model (Fig 1) the wide variation persisted ranging from 3.6% to 57.4% for β-2-agonist s and 0.0% to 65.8% for hypertonic saline. The significant positive association between the proportion of days of hypertonic saline and β-2 agonist use also remained (r² = 0.25, P <.001; Fig 2).

Our sample at the patient level had a median length of stay of 4 days (IQR 2 to 9). Overall, 6.0% of infants had an all-cause readmission within 7 days, 41.6% used an ICU, and 19.2% required mechanical ventilation. In our unadjusted models, there was no significant association between the hospital-level proportion of days of use of hypertonic saline or β-2-agonists and hospital-level clinical outcomes of length of stay, 7-day all-cause readmission, ICU use, and mechanical ventilation use. In our covariate-adjusted model, this absence of association remained (Table 3) between the proportion of days of medication use and each clinical outcome.

In our multicenter study of 6846 children with CHD who were hospitalized with bronchiolitis, there was wide hospital-level variation in the use of β-2-agonists and hypertonic saline. This variation persisted even after adjusting for differences in demographics, complex chronic conditions, and CHD severity among hospitals. There was no association between the hospital-level proportion of days of medication use and clinical outcomes of length of stay, 7-day all-cause readmission, ICU use, and mechanical ventilation use in our adjusted models. Additionally, we found an association between increasing hospital-level β-2-agonist use and hypertonic saline use.

Our finding of variations in the use of β-2-agonists and hypertonic saline mirror variation previously seen in β-2-agonist use otherwise healthy children hospitalized with bronchiolitis^10,11  and supports our initial hypothesis that interventions discouraged for healthy children are being used variably and frequently at some hospitals for children with CHD and bronchiolitis. The management of viral bronchiolitis in infants with CHD is challenging, and decision-making is even more complicated by several factors. Firstly, studies of interventions for bronchiolitis primarily focus on healthy children; therefore, there is a lack of data supporting the appropriate care for this population. Secondly, the higher fragility of this population may lead to a desire to try interventions perceived to be potentially helpful despite the lack of supportive data. The use of these potentially nonbeneficial interventions has known physiologic sequelae and potential side effects such as tachycardia with β-2-agonists²  and bronchospasm with hypertonic saline.^12,13  The strikingly broad variation of the use of β-2-agonists or hypertonic saline in this high-risk yet heterogeneous group should prompt patient-level examination for potential benefits and risks, particularly within CHD subgroups.

The high degree of association between the hospital-level proportion of days of β-2-agonist use and hypertonic saline use in our study that persists even after adjusting for differential disease severity could be influenced by multiple factors. Some sites may have a culture of always using a β-2-agonist in conjunction with hypertonic saline due to concern for hypertonic saline-related bronchospasm, although the evidence for this practice remains controversial.¹²  In the absence of guidelines, some sites may have cultures or practices, such as order sets, that support a more aggressive approach to therapeutic interventions, whereas others are more prone to not intervene and provide supportive care. This finding aligns with other multisite research regarding diagnostic and therapeutic practices in other conditions.^14,15 

There are several limitations to our study. Firstly, PHIS is an administrative database and we relied on billing codes to both identify our population and the use of interventions. Thus, miscoding may have led to inaccuracies. In particular, there may have been systemic differences in coding among hospitals that may have led to under or overestimation of medication variation. Additionally, the PHIS database did not allow us to differentiate between patients who received an intervention once a day versus those who received it multiple times a day. Secondly, although CHD is a heterogeneous group of disorders, we grouped together codes for CHD because it can be challenging to identify specific CHD and stages of surgical correction in administrative databases. We were, therefore, unable to assess whether certain CHD populations might have benefited from interventions. Because of the small number of include infants with single-ventricle disease and potential imprecision in capturing physiology with ICD-9/ICD-10 codes, caution should be used in applying our findings to this group of infants. Thirdly, although we used multiple variables reflecting clinical severity in our models, such as H-RISK score, presence of critical CHD, presence of CHF, and presence of pulmonary hypertension, we may not have adequately adjusted for differences in clinical severity between hospitals and more accurate accounting of this may have decreased the degree of variation in medication use that we noted in our models. Fourthly, because the database only includes the tertiary children’s hospitals that participate in PHIS, our findings are not reflective of patterns of use and outcomes for infants with CHD and bronchiolitis who are cared for at nonfreestanding children’s hospitals, including community hospitals. Additionally, because one of our outcomes was readmission rate, we could not account for readmissions at non-PHIS hospitals and, therefore, may have underestimated this outcome.

In this study to assess intervention use and outcomes in infants with CHD and bronchiolitis, we demonstrate wide hospital-level variations in β-2-agonist and hypertonic saline use in children with congenital heart disease admitted with viral bronchiolitis. These variations are not associated with differences in clinical outcomes, suggesting the potential overuse of these medications in this population. Inhaled medications may be future targets for quality improvement and deimplementation interventions to decrease medication overuse. Additionally, future studies might assess medication interventions in other high-risk medically complex populations that are excluded from the current bronchiolitis guidelines, such as children with chronic lung disease or neuromuscular disorders.