Factors Associated With Bronchiolitis Guideline Nonadherence at US Children’s Hospitals

We used the Pediatric Health Information System (PHIS) database, which includes clinical and resource use data for ED, inpatient, and observation encounters at 52 US children’s hospitals.

We studied bronchiolitis encounters (ED and inpatient and observation admissions) from January 1, 2016 to December 31, 2018, for patients 1 month to 2 years age with an International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) discharge code for bronchiolitis (J21.X) and an All Patient Refined - Diagnosis Related Groups (APR-DRG) for bronchiolitis and respiratory syncytial virus pneumonia (v32; 138).^5,25  Patients seen in the ED before admission were considered as 1 admission encounter. We obtained encounters in the 2 years before our study time frame to determine if a patient had an incident encounter for bronchiolitis before our study period. Encounters occurring within 30 days of an initial encounter were considered revisit encounters and were excluded. Encounters >30 days after an initial encounter were included as secondary encounters. We excluded encounters with any ICU stay, LOS longer than 7 days, or complex medical conditions as defined by Feudtner’s classification²⁶  and/or Pediatric Medical Complexity Algorithm (PMCA).^5,27  We also excluded encounters with concurrent ICD-10-CM diagnosis for croup, pneumonia, or asthma (Supplemental Table 7) because these diagnoses might warrant intervention with any of the 5 metrics.^5,13  This study was deemed exempt from review by our institutional review board.

Our primary outcome, guideline nonadherence, was defined as receiving any of 5 tests or treatments: bronchodilators, chest radiographs, steroids, antibiotics, and viral testing. Each metric was identified by using billing data and the PHIS Clinical Transaction Classification codes: viral testing (test flag “viruses” or “molecular pathology” for named virus), chest radiograph (chest and thorax imaging, technique “X-ray”), antibiotics (therapeutic categories “Aminoglycoside/penicillin” “Cephalosporin/macrolide” “Tetracycline/fluoroquinolone” and “Misc antibiotic/sulfa,” excluding topicals), bronchodilators (therapeutic category “Bronchodilator”), and systemic steroids (therapeutic category “Corticosteroid,” excluding topical/inhalation route). The 5 metrics were dichotomized as either adherent or nonadherent and evaluated as secondary outcomes. All diagnosis codes for patients who received antibiotics were reviewed and categorized by the lead author as “antibiotics indicated” or “antibiotics not indicated” on the basis of expert opinion (see Supplemental Table 3 for specific diagnoses). Antibiotic use was classified as nonadherent only when antibiotics were not indicated by using aforementioned diagnosis code classification.

Patient-level demographic predictors of interest included age, sex, race and ethnicity, insurance, rural or urban residence, median income, illness severity, and medical history. Race and ethnicity were collapsed into a single categorical variable, because >85% of patients who reported being Hispanic and/or Latino ethnicity reported being white or other race. Residence was classified as metropolitan, micropolitan, small town, or rural by using the Rural-Urban Community Area codes provided by PHIS.²⁸  Median income was assigned by using the estimated median income for a patient’s zip code. Illness severity was defined by using APR-DRGs combined into 3 categories: minor, moderate, and major or extreme.^5,29  Patients with noncomplex chronic conditions were identified by using PMCA.²⁷ 

Contextual factors included transfer from outside facility (as classified by PHIS, included transfers from any health care facility type), discharge quarter, and whether it was a patient’s incident or secondary encounter.

Provider characteristics were only included in ED analyses, because most encounters (94%) had 1 attending provider, allowing us to attribute guideline nonadherence. In contrast, inpatient encounters had multiple providers, making attribution challenging. Provider specialty in the ED was classified as pediatric emergency medicine (PEM), pediatrics, certified nurse practitioner or physician’s assistant (CNP/PA), or other. We excluded 1.3% of ED encounters that had no provider listed. Census region was included as a hospital-level factor.

We used descriptive statistics to summarize bivariate associations of nonadherence with demographic and contextual factors.

For the primary analysis, we used mixed effects logistic regression for binary nonadherence outcomes. All factors were included as fixed effects and summarized with adjusted odds ratios (aORs) and 95% confidence intervals (CIs). Odds ratios >1 indicate higher nonadherence; analogously, odds ratios <1 indicate better adherence. The model included hospital random intercepts and, for ED encounters, provider random intercepts, which account for correlation within hospital and provider. We used intraclass correlation coefficients (ICCs) to summarize the proportion of variation in nonadherence explained by hospital and provider.

For all analyses, we used Stata (Stata Corp, College Station, TX) and SAS (SAS Institute, Inc, Cary, NC), specifically the glimmix procedure from SAS for regression models.³⁰ 

A total of 198 028 encounters met inclusion and exclusion criteria: 141 442 ED and 56 586 admitted (Fig 1). There was nonadherence in 46.1% of encounters: 40.2% among ED encounters and 61.0% among admissions (Table 1). Bronchodilators (23.1%) and chest radiographs (20.7%) were the most frequently ordered tests and treatments. Antibiotics were indicated for concurrent diagnoses in 68.2% of patients who received them (see Supplemental Table 3).

Nonadherence increased with patient age, with both ED and hospital providers being more likely to order tests and treatments for children 12 to 24 months compared with infants 1 to 2 months (ED: aOR, 3.39; 95% CI, 3.20–3.60; admissions: aOR, 2.97; 95% CI, 2.79–3.17]). Higher systemic steroid use (aOR, 45.69; 95% CI, 29.26–71.34 in ED and aOR, 32.15; 95% CI, 25.62–40.33 for admissions) drove this difference (see Table 2; Supplemental Tables 3 and 4; and Figs 2 and 3).

Nonadherence was higher for admitted non-Hispanic Black patients (aOR, 1.16; 95% CI, 1.10–1.23) compared with non-Hispanic white patients, driven by higher rate of steroids (aOR, 1.29– 95% CI, 1.17–1.41) and bronchodilators (aOR, 1.39; 95% CI, 1.31–1.48) but not antibiotics (aOR, 0.73, 95% CI, 0.66–0.81) (Fig 3; Supplemental Table 6). Admitted American Indian patients were more likely to receive guideline adherent care compared with non-Hispanic white patients (aOR, 0.67; 95% CI, 0.48–0.95), driven by low receipt of bronchodilators (aOR, 0.56; 95% CI, 0.36–0.87) and chest radiographs (aOR, 0.63; 95% CI, 0.42–0.95).

Compared with patients from an urban zip code, patients living in micropolitan (aOR, 0.68; 95% CI, 0.63–0.73), small town (aOR, 0.68; 95% CI, 0.61–0.75), and rural (aOR, 0.65; 95% CI, 0.57–0.75) zip codes were more likely to receive guideline adherent care during an admission encounter even after adjusting for transfer status. (Supplemental Table 6).

Although overall, guideline nonadherence was similar between provider types, compared with PEM physicians, CNP/PAs were more likely to order viral testing (aOR, 1.61; 95% CI, 1.32–1.95), but less likely to prescribe bronchodilators (aOR, 0.47; 95% CI, 0.39–0.57), steroids (aOR, 0.47; 95% CI, 0.37–0.60), or antibiotics (aOR, 0.16; 95% CI, 0.07–0.38).

Variation in hospital (random) effects explained the most variation in viral testing in both the ED (ICC: 0.35) and admission (ICC: 0.32) encounters and was least prominent for antibiotic use (ED ICC: 0.12, admission ICC: 0.02). Variation in ED provider (random) effects explained an additional percentage of the variation (Supplemental Table 5). For example, the ICC for viral testing and antibiotics increased to 0.49 and 0.18, respectively.

We found multiple factors associated with nonadherence to the most recent AAP bronchiolitis guideline including age, race and ethnicity, and provider specialty. In addition, for several metrics such as viral testing, there was variation across hospitals and providers. Variation may signal an opportunity to reevaluate best practice, clarify guideline recommendations, and perform targeted deimplementation to achieve the desired level of adherence.

Previous work set ABCs for bronchiolitis guideline nonadherence in ED²  and hospitalized³  patients by ranking the best-performing PHIS hospitals comprising at least 10% of the total population (eg, top 4–6 hospitals) and calculating a median of those top performers. This approach supports the notion that some deviation from guidelines is likely appropriate, patient-sensitive variation. Among the ED cohort, rates of nonadherence were comparable with the 2005 ABC for chest radiographs (16.7% vs 17% ABC) and antibiotics (0.6% vs 1.85% ABC).²  For admission encounters, rates were also similar or better than 2012 ABCs for chest radiographs (31.2% vs 32.4% ABC) and antibiotics (7.9% vs 18.5% ABC) but not for viral testing (21.6% vs 0.6% ABC) or steroids (9.7% vs 6.4% ABC).³  We did not directly compare bronchodilator ABCs because of different study definitions.³  With our study, we suggest that PHIS hospitals are achieving several of the bronchiolitis ABCs, but deimplementation opportunity remains for viral testing and steroid use.

Several patient-level factors were found to be drivers for nonadherence. Younger patients were more likely to have viral testing, perhaps in part influenced by evaluation of fever in patients 1 to 2 months of age.³¹  Older patients, in whom asthma is more commonly considered, were more likely to receive bronchodilators and steroids.³² 

Race and ethnicity were also associated with differences in adherence. Non-Hispanic Black patients were less likely than non-Hispanic white patients to receive antibiotics, which has also been seen in other investigations.^33,34  Similar to 1 previous study, we found that non-Hispanic Black patients were more likely to receive bronchodilators than other groups.²⁰  Providers may be more likely to order bronchodilators in non-Hispanic Black children because of higher population asthma prevalence or family history.³⁵  In contrast, American Indian patients had a 50% lower odds of receiving bronchodilators than non-Hispanic white patients despite previous studies revealing higher asthma prevalence in American Indian and Alaskan Native patients.^36,37  Although clinical decision-making cannot be determined in our study, there may be many factors that explain racial differences, including how patients present for medical care, social inequities, parental expectations for tests or treatments, institutional racism, and provider implicit bias.^20,38–46 

Admission encounters from lower-population density areas were more likely to receive guideline adherent care than those from urban areas. Reasons why patients from rural zip codes received less additional tests and treatments are unclear but may relate to care received before arrival at a children’s hospital. Although we account for transfers in our regression model, it is possible that patients received pre-encounter care not captured in our study.

Higher use of bronchodilators and steroids were noted during secondary encounters. Researchers in numerous studies have explored the natural history of airway hyperresponsiveness, recurrent wheeze, and asthma following bronchiolitis episodes.^47–53  Providers may be more likely to trial bronchodilators and steroids in children with previous bronchiolitis because of perceived potential for benefit.

We found provider type to be a factor associated with differences among ED encounters. Pediatricians and CNP/PAs were more likely to obtain viral testing but less likely to give antibiotics, bronchodilators, or steroids than PEM physicians. Provider beliefs, knowledge, and social influence contribute to bronchiolitis management practices.^14,54,55  CNPs/PAs may care for lower acuity patients, for example, children without respiratory distress where interventions (eg, bronchodilator trial) are less likely. However, we attempted to adjust for this by excluding encounters with any ICU care and adjusting for severity of illness in the regression model.

Researchers in previous studies have found varying degrees of nonadherence to AAP bronchiolitis guidelines across hospitals.^4–6  We found significant hospital-level variation for several guideline metrics, most notably viral testing. The AAP guidelines do not discourage against influenza testing, which may be included in viral testing panels.¹  Local guidelines may include scenarios for obtaining viral testing, such as in younger patients, prolonged illness, influenza risk factors, or for infection control measures.^22,56 

The potential impact of bronchiolitis guideline nonadherence is broad. Patients exposed to unnecessary interventions may experience overdiagnosis and increased LOS and costs, as well as drug intolerance or reaction, distress from blood draws, radiation exposure, and downstream effects such as increased risk of obesity associated with antibiotic use.^57–63 

The impact on health systems is another important consideration. In previous studies, researchers found that use of antibiotics, bronchodilators, chest radiographs, and corticosteroids were associated with longer hospital LOS with no reduction in readmission.^5,8  In contrast, higher guideline adherence has been associated with reduced LOS and cost.¹³ 

Although multiple studies have revealed strategies for reducing bronchiolitis guideline nonadherence by using quality improvement methodology,^4,10,64–70  our study reveals that there is ongoing opportunity for improvement at US children’s hospitals. A guideline’s success depends on provider knowledge and belief in the application of recommendations to their individual patient. We identify specific factors that are associated with nonadherence, which may inform local and national deimplementation initiatives.

We also suggest there may be opportunities to advance health equity in bronchiolitis care by providing the right resources, rather than the same resources, to different patients. We found a higher use of bronchodilators and steroids among non-Hispanic Black patients but not among American Indian patients, despite both of these populations having higher asthma prevalence than non-Hispanic white patients. The AAP guidelines acknowledge limitations, including heterogeneity and small sample size, to studies informing bronchodilator recommendations.¹  In addition, studies did not perform any subgroup analysis to determine if there was potential benefit within certain populations such as those with higher reported lifetime rates of asthma.^71–73  Further research into the outcomes for diverse groups of patients trialing bronchodilators in bronchiolitis is warranted. Future AAP guideline iterations may consider addressing evidence and tailoring recommendations to different patient populations, such as those with higher asthma prevalence.^74–76 

This was a retrospective administrative database study, so provider reasoning was not included. The PHIS database does not identify where the intervention occurred, thus for admitted patients metrics may not reflect inpatient provider behavior. The PHIS database includes only large children’s hospitals, and applicability of this study to other settings is unknown. We used APR-DRG severity of illness for adjustment because this is the tool available in PHIS; however, its utility in assigning severity of bronchiolitis is unclear. For this reason, we also excluded encounters with any ICU care and separated analysis into ED discharges and admitted patients to account for other severity indicators. Collection for race and ethnicity data may vary across hospitals, and categorization validity have been previously questioned.⁷⁷  Patient primary language, a driver for guideline nonadherence, was not included.²¹  We did not include discharge prescriptions, which may overestimate adherence for some metrics. Because we were unable to determine presence or content of local guidelines, we described nonadherence to nationally available guidelines, which may differ from local recommendations. We focused on factors that may contribute to guideline nonadherence; researchers in subsequent studies may continue to examine the impact of nonadherence on outcomes.

Multiple patient, provider, and hospital factors are associated with nonadherence to national bronchiolitis guidelines. Tailoring guidelines and benchmarks for different populations on the basis of disease prevalence and risk factors may lead to more targeted practice.