Variation in Dexamethasone Dosing and Use Outcomes for Inpatient Croup

This retrospective cohort study of children admitted with a primary diagnosis of croup was part of a larger previously published prospective cohort study across 5 Pediatric Research in Inpatient Settings Network children’s hospitals in the United States between July 2014 and June 2016.¹⁷  Children were eligible for enrollment if they were 6 months to 6 years of age and the family spoke either English or Spanish. Children were not eligible if they had the following comorbidities: congenital lung or airway disease, neuromuscular disease, congenital heart disease, immunodeficiency syndromes, cancer, and sickle cell disease. Recruitment and enrollment procedures are detailed in a previous publication.¹⁷  In the parent study, enrollment of children admitted with croup was prioritized, but children seen and discharged from the emergency department (ED) were also enrolled. For this study, only admitted patients were included in the primary analysis, but patients discharged from the ED were included in a secondary analysis, as described in Supplemental Fig 1.

After enrollment, patients were excluded from the analyses if they received steroids other than dexamethasone (eg, prednisone, methylprednisolone) by the intravenous or oral route during their ED visit or admission. This was done to account for the differing properties of dexamethasone compared with other steroids in terms of potency, half-life, and dosing frequency. If a child met inclusion criteria for enrollment, but their discharge diagnosis was not croup, they were excluded from the analysis. Children were excluded if they were admitted or transferred to the ICU. Thirteen patients were excluded from the primary analysis because we were unable to categorize them into a dexamethasone dosing category. This included 11 children with a record of dexamethasone administration during (but not before) their ED visit or hospitalization per medical record (MR) data abstraction but no record of dexamethasone administration in PHIS and 2 patients without any record of dexamethasone administration. We performed a sensitivity analysis that included these 13 excluded patients.

The institutional review boards at each participating hospital and the Western Institutional Review Board approved this study.

In this study, we employed 3 data collection methods. Data were collected by using the PHIS database (Children’s Hospital Association, Overland Park, KS), MR abstraction, and questionnaires completed by parents during the index ED visit or hospital admission and after discharge.¹⁷ 

The PHIS database was used to collect data on specific moderating factors that we hypothesized may influence outcomes of interest, including child age, sex as a biological variable, level of medical complexity, insurance type, visit season, discharge location (ED, inpatient), and outcomes, including LOS, charges, and 30-day same-cause reuse (repeat ED visits or readmissions for a primary diagnosis of croup). PHIS was also used to collect the number of dexamethasone administrations after presentation to the participating hospital or ED and total hospital charges.

Two trained team members at each participating hospital abstracted data form the MR of the participating hospital. Abstractors trained until their abstractions aligned with gold standard abstractions, with a κ statistic of ≥0.75.¹⁷  MR abstraction was used to verify whether children received dexamethasone in the 24 hours before presentation to the participating hospital (eg, other hospital, during medical transport, physician office). Abstractors categorized patient severity before the first pharmacologic treatment as mild, moderate, or severe on the basis of a priori established criteria.¹⁸  Severe classification was assigned if patients were described as severe by the clinician: had stridor and/or retractions at rest and lethargy or agitation, were treated with racemic epinephrine and were agitated or lethargic, or had impending or actual respiratory arrest or failure. Moderate classification was assigned if children were described as moderate by the clinician: had stridor and/or retractions at rest but were not lethargic or agitated or were treated with racemic epinephrine but were not agitated or lethargic. Mild classification was assigned if children were described as “mild” by the clinician: had no significant stridor or no stridor or significant chest wall retractions at rest and no lethargy or agitation. When documentation supported >1 severity classification, patients were assigned the highest supported severity classification. Patients were dichotomized as mild and moderate or severe for analyses. Questionnaires completed by parents during the index visit collected child race and ethnicity and caregiver education. The Pediatric Medical Complexity Algorithm (PMCA) was used to classify children into 3 categories of medical complexity: (1) no chronic illness, (2) noncomplex chronic illness, or (3) complex chronic illness.¹⁹ 

The predictor variable was dexamethasone dosing regimen category (1 vs ≥2 doses of dexamethasone). Patients were categorized into dosing groups by using the total sum of dexamethasone doses received. This included doses received 24 hours before presentation to the participating hospital plus all doses received during the index ED visit and admission.

Outcome variables were as follows: (1) LOS, (2) cost, and (3) hospital reuse (ED revisits or readmissions). Total hospital charges were converted to costs by using institutional cost-to-charge ratios. Costs were adjusted for inflation to 2018 dollars by using the medical care services component of the Consumer Price Index.²⁰  Costs were then winsorized at the fifth and 95th percentiles to reduce the effect of extreme cost outliers.²¹ 

Analyses were performed at the patient level. We examined univariate descriptive statistics for all predictor and outcome variables. χ² Tests were used to examine the relationship between dexamethasone dosing category and covariates and outcomes. To examine differences across sites, we tabulated LOS in hours and the number of admitted patients in each dexamethasone dosing regimen category (1 vs ≥2 doses of dexamethasone) by hospital.

Multivariable generalized linear and logistic mixed-effects regression models were used to examine the relationship between dexamethasone dosing group and outcomes. For cost, we applied linear models. For LOS, we applied a log-γ model. For reuse, we applied a logistic model. In all multivariable-adjusted analyses, we included the following covariates, which were selected a priori: child age, child sex, child race and ethnicity, severity on presentation, PMCA category, insurance type, caregiver education, and hospital. In the multivariable-adjusted analysis of cost, we also controlled for LOS because LOS is the main driver for inpatient cost. A site-specific random effect was also included in all the models to account for clustering within hospital. We compared 30-day same-cause ED and/or inpatient reuse at the original site of care across dexamethasone dosing categories. This outcome was binary; each patient was classified as having or not having reuse. This outcome was too rare to produce reliable coefficient estimates when all covariates were included in the model, so a reduced regression model that did not include PMCA or site had to be applied.

To understand differences between children receiving and those not receiving steroids before ED or hospital arrival, we first compared patient characteristics (severity on presentation category, admission versus discharge from ED) between children who received dexamethasone before ED or hospital arrival (n = 91) and children who did not (n = 208). Patients seen in the ED and discharged (n = 63) and admitted patients (n = 234) were included in this portion of the secondary analysis. Second, for children hospitalized with croup, we compared LOS and cost between children who received dexamethasone before ED or hospital arrival (n = 82) and children who did not (n = 152) in an unadjusted analysis using a Kruskal-Wallis test to compare medians for these skewed data.

All analyses were conducted by using R version 4.0.2.¹⁷ 

Among 234 hospitalized children, 145 (62%) received 1 dose and 89 (38%) received >1 dose. Of those who received >1 dose, 62 (70%) received 2 doses, 19 (21%) received 3 doses, and 8 (9%) received ≥4 doses. Patient characteristics, including severity on presentation, did not differ statistically across dexamethasone dosing regimen groups (Table 1). Those who were considered severe at presentation were less likely to receive >1 dose of dexamethasone than those with mild or moderate symptoms (31.9% vs 40.2%), but this difference was not statistically significant. The proportion of admitted children who received >1 dose varied across hospitals, ranging from 27.9% to 57.1% (P = .04) (Table 2).

LOS was longer for children who received >1 dose of dexamethasone (43.7 vs 27.0 hours; P ≤ .001). In the adjusted log-γ regression, receiving >1 dose of dexamethasone was associated with 45% longer LOS compared with receiving only 1 dose of dexamethasone (relative risk [RR] = 1.45; 95% confidence interval [CI]: 1.30 to 1.62). Hospital 2 prescribed >1 dose to 52.6% of admitted patients and had a statistically significant longer LOS (RR = 1.57; 95% CI: 1.34 to 1.85) compared with hospital 1. None of the other hospitals had statistically significant differences in adjusted LOS compared with hospital 1 (Table 3).

In the unadjusted analysis, receipt of >1 dose of dexamethasone was associated with increased cost compared with receipt of 1 dose ($4157 vs $2720; P ≤ .001) (Table 4). In the adjusted linear regression without LOS, receiving >1 dose of dexamethasone was associated with a $960 higher cost (95% CI: $455 to $1464). However, after we controlled for LOS, the association between >1 dose of dexamethasone and cost, compared with 1 dose, was not statistically significant ($−31.2; 95% CI: $−424 to $362) (Table 3). One extra day of LOS was associated with a $1712 increase in cost (95% CI: $1471 to $1953). Thirty-day same-cause reuse was rare (n = 18, 7.7%). There was no statistically significant association between number of dexamethasone doses and 30-day same-cause reuse (odds ratio: 0.87; 95% CI: 0.26 to 2.95).

When the 11 patients with conflicting steroid data and the 2 patients without dexamethasone were included in a 0-dose dexamethasone dosing category in the multivariable linear and logistic mixed-effects regression models, the associations between dosing category (0, 1, and >1) and outcomes (LOS, cost, and reuse) did not change.

Ninety-one (30%) of the 299 patients seen and discharged from the ED or admitted had documentation of receiving dexamethasone before ED or hospital arrival. Whether patients received dexamethasone doses before arrival to the ED or hospital did not differ statistically across severity on presentation categories (P = .25), but children who received dexamethasone before arrival were more likely to be admitted (90.1% vs 74.0%; P = .002). Among admitted patients who received dexamethasone before arrival, 45 (55%) received >1 dose (including doses received 24 hours before and during their index ED visit or admission). Two-group comparisons using the Kruskal-Wallis test revealed that admitted patients who did receive dexamethasone before ED or hospital arrival had a higher unadjusted cost ($3595.70 [SD $2384.11] vs $3009.25 [$1958.63]; P = .048) but no difference in hours of LOS (35.12 [SD 22.37] vs 32.14 [17.88]; P = .26) compared with admitted patients who did not receive dexamethasone before ED or hospital arrival.

We found statistically significant variation in the proportion of children receiving multiple dexamethasone doses across participating hospitals. This study aligns with a previous retrospective study that found significant variation across hospitals in the use of multiple doses of dexamethasone for hospitalized children with croup.¹  Importantly, severity on presentation was not associated with receiving multiple doses of dexamethasone, suggesting that the variation across hospitals may be related to other factors, including worsening or returning symptoms during hospitalization, provider perception that multiple doses improve symptoms or prevent return of symptoms, provider or parent preference, care pathway recommendations based on local expert consensus, and/or institutional culture.

Multiple doses of dexamethasone was associated with a 45% longer LOS. Prolonged LOS meaningfully increases cost. A single extra inpatient day increased cost, on average, by $1712. Some clinicians may routinely keep children hospitalized to give more dexamethasone doses. Both the proportion of children receiving multiple doses and LOS varied across hospitals. Interestingly, hospital 2 prescribed >1 dose of dexamethasone to 53% of admitted children, compared with 40% at hospital 1, and had a 58% longer adjusted LOS. Care pathway recommendations based on local expert consensus and/or institutional culture might have contributed to these differences. Conversely, it is possible that differences in LOS and dexamethasone prescribing are related to ongoing symptoms (such as respiratory distress) in a subset of children influencing clinicians to delay discharge and prescribe additional dexamethasone doses. Although reuse rates were overall higher in our study than in those previously published,¹  reuse did not differ between dosing groups, which may indicate that clinicians appropriately redose dexamethasone in some children.

In this study, we explored differences between admitted children who received steroids before ED or hospital arrival and those who did not. Among admitted children, approximately one-third of children (35%) received steroids before arrival. Presumably, these children had persistent symptoms prompting presentation to the ED. However, 45% did not receive any additional steroids in the ED or hospital. In other words, among all admitted children in the study, 26% of children in the 1 dose group received their 1 dose of dexamethasone before presentation. This aligns with previously published PHIS data that revealed 19% of patients did not receive any steroids in the ED or hospital.¹  Receiving steroids before arrival did not shorten LOS (35.12 [SD 22.37] vs 32.14 [SD 17.88]; P = .26) but was associated with a higher unadjusted cost of admission ($3595.70 [SD $2384.11] vs $3009.25 [SD $1958.63]; P = .048). It is possible that children admitted after initial outpatient treatment were perceived as atypical, prompting longer observation periods before discharge. When including both children discharged from the ED and those admitted, receiving dexamethasone before presentation was not associated with severity on presentation but was associated with higher admission rates (90.1% vs 76%; P = .004). Further research is needed to understand drivers of failed outpatient management and admission for croup.

Croup can cause anxiety and concern for both parents and providers, which may drive the practice of prescribing multiple doses of steroids for croup. However, steroids are known to be associated with undesired side effects, such as hyperactivity, emotional lability, sleep disturbance, and, rarely, secondary bacterial infection.^13–16  Given the long half-life of dexamethasone, these symptoms may persist for several days after administration and may be compounded by administration of multiple doses. A prospective randomized controlled noninferiority trial comparing 1 versus multiple doses of dexamethasone for hospitalized children with croup is needed to balance benefits and potential harms and inform best practice. In such a trial, researchers should measure and adjust for severity of symptoms beyond ED triage, anticipating issues of residual confounding that simple randomization may not surmount, and incorporate outcomes important to parents and patients, such as untoward medication side effects and the ability to return to day care or school.

These findings should be interpreted in light of the study strengths and limitations. This was an observational study; we report associations and may not have accounted for all relevant confounders. We collected and controlled for severity before receipt of interventions using chart abstraction. However, given limitations of the MR, our severity classification may be incomplete or inconsistent across sites or clinicians.²²  Severity was only collected on presentation. For children who received steroids before presentation, the severity assessment on presentation was likely affected by the previous treatment. Also, interestingly, children categorized as severe at presentation were less likely to receive >1 of dexamethasone compared with children categorized as mild or moderate. We did not abstract ongoing assessments of severity as the hospitalization progressed. It is possible that severity on presentation may not be predictive of future illness course and that worsening severity during the hospitalization partially explains differences in dexamethasone dosing and LOS. Although we used chart abstraction to capture whether children received steroids before presentation to the study hospital, the actual dose of steroid was not captured, nor was the dose relative to each child’s weight. Although we were able to compare the number of doses, we cannot determine if each dose was the typical dose of 0.6 mg/kg. It may be that children who received a lower dose before ED presentation were more likely to receive a second dose of dexamethasone. More granular, contextual information is needed to identify factors that lead clinicians to redose dexamethasone. We only captured reuse at the participating hospital, which may underestimate actual reuse. We examined dexamethasone use across 5 children’s hospitals. Our findings may not be generalizable to all settings given that all hospitals in our study were university-affiliated freestanding children’s hospitals and many children with croup are likely seen in community hospitals.²³ 

In this multisite prospective study of children hospitalized with croup, we found statistically significant variation in dexamethasone dosing independent of severity on presentation. Patients receiving multiple doses of dexamethasone had longer LOS but no differences in adjusted cost or rates of repeat ED visits or readmissions compared with patients receiving 1 dose. A randomized trial is needed to compare 1 versus multiple doses of dexamethasone for hospitalized children with croup, and additional research is also needed to understand drivers of failed outpatient management and admission for croup.