Dexamethasone Versus Prednisone in Children Hospitalized With Asthma Exacerbation

We conducted a retrospective cohort study at an urban, quaternary children’s hospital from January 1, 2014 to December 31, 2017. Children with asthma exacerbations receiving continuous albuterol or intermittent dosing as frequently as every 1 hour are admitted to general inpatient units at our hospital.

We included all children 4 to 17 years old hospitalized with a primary discharge diagnosis of asthma (International Classification of Diseases, Ninth Review Clinical Modification [ICD-9-CM] 493.XX or International Classification of Diseases, 10th Review, Clinical Modification J45.xx). We excluded children with medical complexity by using International Classification of Diseases codes consistent with tracheostomy, cystic fibrosis, bronchopulmonary dysplasia, and congenital heart disease.¹⁰  We only included children treated with both a systemic corticosteroid and an inhaled bronchodilator, to select patients with an acute asthma exacerbation and not children with asthma admitted for another indication. We excluded children admitted for anaphylaxis or croup (because they may have received both steroids and a bronchodilator), those whose presentation was severe enough to initially require admission to the PICU, and those who left against medical advice or transferred to another hospital.

Most data elements were abstracted from the electronic health record; however, a medical record review was completed by 3 trained reviewers to detail receipt of previous admission medications, including antibiotics, inhaled corticosteroids, antihistamines, leukotriene modifiers, and systemic steroids. We assured interrater reliability by calculation of Fleiss’s κ statistic on 10% of the charts.

To avoid issues of nonindependence, children with multiple admissions during the study period had a single admission randomly chosen for inclusion in the study cohort.

The hospital’s institutional review board approved this study.

The primary predictor was the first systemic corticosteroid, dexamethasone compared with prednisone and prednisolone (PRED), administered after presentation to our ED or inpatient unit for acute asthma exacerbation. A secondary analysis was performed examining treatment with dexamethasone only versus PRED only for the entire hospital steroid course. Steroids administered before arrival to the hospital were not considered in these analyses as we focused on therapies administered during the encounter. Since steroids administered before arrival may influence treatment decisions, this covariate was included in the propensity model discussed below.

Our primary outcome was asthma-specific health care reutilization including urgent care, ED, or unplanned hospital readmission¹¹  with a principal diagnosis consistent with asthma (as defined above) within 30 days of the index hospital discharge. Secondary outcomes included the index hospitalization length of stay and the rate of decompensation requiring transfer to the PICU after initial admission to a general inpatient unit. Length of stay was calculated as the time elapsed, in hours, from presentation to our ED or inpatient unit until hospital discharge.

Because physicians consider a constellation of clinical factors in choosing a steroid for asthma exacerbations, we used a covariate balancing propensity score in an inverse probability treatment weighting (IPTW) model to account for confounding in the physician discretion of steroid selection.¹²  All covariates were determined a priori on the basis of existing literature and clinical knowledge from the study team (Tables 1 and 2).⁷  The propensity score was created independent of the outcome measures. The propensity score balance was examined by plotting the standardized mean difference between two groups both before and after the IPTW. Covariates with standardized mean difference <0.2 were considered balanced.

For outcome models, we used generalized linear models with appropriate link functions, including the propensity score weights and relevant in-hospital covariates to compare the treatment effectiveness by steroid treatment groups (Table 2). Given all pretreatment covariates in the propensity score model were balanced after the application of the IPTW, these covariates were not included in the outcome model. A logit link function was used for the 30-day return utilization (urgent care, emergency visit, or unplanned readmission) and transfer to the PICU. Because of a nonnormal distribution, length of stay was log-transformed before modeling by using a gaussian link function. After modeling, the length of stay results were back-transformed before presenting results. Before establishing our final model, we evaluated covariate candidates over time to ensure that the evolving clinical care practices at our institution during the study period would not bias our results (Supplemental Information).

Based on historical utilization, our a priori power calculation assumed that our 3-year study period would have 800 patients in each treatment cohort and a 7% return utilization. Thus, we would be powered to detect a 4.5% difference in return utilization with a power (β) of 0.9.

Our propensity score model included clinical factors available to the provider at the time of steroid selection. This model included 3 categories of covariates: demographic and clinical variables, markers of acute illness severity, and markers of chronic asthma severity (Table 1).

Demographic and clinical characteristics included age, sex, race, weight greater than 85th percentile for age, payer type, and the census tract level poverty rate of the reported home address. We included patient race in the propensity score model as physicians may have explicit or implicit biases in prescribing medications. Census tract-level poverty was obtained from the US Census Bureau’s 2011–2015 American Community Survey after the assignment of an individual census tract from a geocoded patient address.^13,14 

Markers of acute illness severity included abnormal vital signs at presentation, diagnostic tests or therapeutic interventions performed within the first 5 hours of presentation, and the prehospital administration of a systemic corticosteroid, all of which may have influenced the provider’s steroid treatment selection. Abnormal vital sign parameters were as follows: heart rate and respiratory rate greater 95th percentile for age,¹⁵  hypoxia as defined by pulse oximetry reading less than 90%, and presence of fever. Diagnostic or therapeutic interventions included completion of a chest radiograph, administration of continuous albuterol, antibiotics, or intravenous magnesium.

Markers of chronic disease management included prescription of inhaled corticosteroid (with or without long-acting β agonist) before admission and asthma-specific hospitalization occurring in the 12 months preceding index hospitalization.

Reutilization and our secondary outcomes are potentially impacted by many factors in addition to the initial steroid. Therefore, the outcome models were adjusted for season of presentation and in-hospital asthma management decisions that occurred after the initial corticosteroid treatment, including initiation of a new adjunctive asthma medication (leukotriene modifier, oral antihistamine, or inhaled antihistamine) and initiation or step-up in an inhaled corticosteroid (Table 2). The step-up in inhaled corticosteroid was determined by using age-specific categories of low, medium, and high daily dosage classification of daily inhaled corticosteroid dose with or without a long-acting β agonist per the National Institutes of Health, National Heart, Lung, and Blood Institute Expert Panel Report 3 Clinical Practice Guidelines.¹⁶ 

A dataset of 20% of charts (n = 239) was completed by all reviewers with responses having a Fleiss’s κ statistic assessing for interrater reliability among the 3 reviewers of 0.895.

Of the 3490 hospitalizations for asthma during the study period, 1694 met the inclusion criteria (Supplemental Fig 2). A total of 259 patients had >1 index encounter during the study period, for a total of 647 encounters (range 2–14 admissions). To avoid confounding due to nonindependence, 1 index admission per patient was randomly chosen and 388 additional encounters were omitted, after which 1306 patients (each with a single index encounter) remained in the cohort. An additional 101 patients were excluded for an initial treatment with methylprednisolone, and 44 patients were excluded for missing covariate values. The primary analysis cohort, based on first treatment dose, included 1161 patients. The secondary analysis, treated with a single corticosteroid medication, excluded patients who receive >1 systemic corticosteroid medication during their admission, for a total of 1062 patients.

Dexamethasone was administered as the first systemic corticosteroid to 510 (43.9%) patients and 651 (56.1%) patients were treated initially with PRED (Table 3). There was no difference in mean age among children treated with a first dose of dexamethasone (8.4 [3.4] years) versus PRED (8.5 [3.4] years). Most patients in the cohort were male (59.2%) and nonHispanic black (50.0%), whereas only 29.9% had private insurance. A third (34.9%) of patients had weight >85th percentile for sex and age. The mean census tract fraction of poverty for the cohort was 24.8 ±18.8%. For context, in 2017, 13.4% of the US population lived in poverty; any census tract with a poverty rate ≥20.0% is a poverty area.¹⁷  There was no statistical difference by group for age, sex, race, weight >85 percentile, payer type, or home census tract poverty rate in the unadjusted cohort.

On presentation, children initially treated with dexamethasone were more likely to have a heart rate >95th percentile for age and sex (45.0%) compared with PRED (36.9%, P = .007). There was no difference by treatment of a respiratory rate >95th percentile, fever, or hypoxia on presentation. Children initially treated with dexamethasone were more likely to be treated with intravenous magnesium (9.8%) compared with PRED (3.1%, P < .001). There were no differences by treatment of the remaining diagnostics or interventions in the first 5 hours after presentation, including obtaining a chest radiograph, initiation of continuous albuterol, or treatment with antibiotics. Children initially treated with dexamethasone were less likely to have been prescribed an inhaled corticosteroid before admission (38.2%) compared with PRED (49.0%, P < .001). There was no difference in the 2 groups in rates of previous asthma hospitalization in the preceding 12 months. Prehospital systemic corticosteroids were given to 23.1% of patients; PRED was the most common prehospital steroid treatment (68.7%) whereas dexamethasone was the next most common (20.9%). Those who received a prehospital corticosteroid (dexamethasone, PRED, or other) were noted to have significant differences in their first hospital-based corticosteroid treatment (Table 3).

The covariate-balanced cohort, reflecting the application of the IPTW to balance differences in the covariates, included the same 1161 patients as the full cohort, including 510 (43.9%) treated with dexamethasone and 651 (56.1%) who were first treated with PRED. All covariates were balanced in which the maximum standardized mean difference was −0.11 (Fig 1). There were no differences in demographic characteristics, markers of acute illness severity, or markers of chronic disease management in the balanced cohort (Table 3).

Additional covariates included in the outcome model were similar between the children who received dexamethasone and PRED as their first hospital-administered steroid dose, with minor statistical differences in the season of presentation (Table 3).

The 30-day return utilization in the cohort was 2.9% (n = 34). In adjusted analyses, the odds of unplanned reutilization associated with receiving a first dose of dexamethasone compared with PRED were 1.61 (95% CI, 0.80–3.31). There was no difference in adjusted odds of transfer to the PICU (dexamethasone 1.6% versus PRED 1.4%; OR 1.21; 95% CI 0.44–3.23) or in the median length of stay (dexamethasone 31.0 [IQR 23.5–43.1] hours versus PRED 31.1 [IQR 23.2–39.8] hours; OR 1.03; 95% CI 0.98–1.08) (Table 4).

In the secondary analysis comparing the group of patients treated with only dexamethasone versus only PRED during their index encounter, dexamethasone was administered as the only systemic corticosteroid to 452 (42.6%) patients whereas 610 (57.4%) patients were treated with only PRED, (Supplemental Fig 2). There were no differences in demographic characteristics between the 2 groups (Supplemental Table 5). Differences between the dexamethasone only and PRED only groups are similar to the differences in the first-dose steroid cohorts. Specifically, children who received dexamethasone only had higher presenting heart rates; more were treated with magnesium, but reported inhaled corticosteroids as a controller medicine less frequently than the PRED only group (Supplemental Table 6).

After the application of the IPTW, the covariate-balanced cohort included 1062 patients, including 452 (42.6%) treated with dexamethasone and 610 (57.4%) treated with PRED. There were no differences in demographic characteristics, markers of acute illness severity, or markers of chronic disease management in the balanced cohort (Supplemental Table 6). The standardized mean differences for the full cohort and covariate-balanced cohort after application of the IPTW are reflected in Supplemental Fig 3.

Additional covariates included in the outcome models were similar between the children who received full courses of either dexamethasone or prednisone, with statistical differences in the season of presentation (Supplemental Table 6).

The overall 30-day return utilization in the full course steroid cohort was 3.1% (n = 33). The adjusted odds of unplanned reutilization associated with receiving only dexamethasone compared with PRED were 1.71 (95% CI, 0.84–3.54). There was no difference in adjusted odds of transfer to the PICU (dexamethasone 0.9% vs PRED 0.5%; OR 1.88; 95% CI 0.39–10.08) or in the median length of stay (dexamethasone 29.1 [IQR 22.8–39.9] hours versus PRED 29.9 [IQR 22.8–38.7] hours; OR 1.00; 95% CI 0.95–1.05) (Table 5).

Our single-center, retrospective cohort study compares outcomes with dexamethasone versus prednisone for the management of children hospitalized for an asthma exacerbation. Our analysis using propensity score methodology for causal inference found no significant differences in the outcomes of asthma-specific 30-day return utilization, index length of stay, or rate of transfer to the PICU by steroid treatment. These results are consistent with previous reports,^5,7–9  and strengthen our understanding of the routine use of dexamethasone in children hospitalized for asthma exacerbations; our analyses incorporated patient-level details through propensity weighting to minimize confounding due to differences in acute severity on presentation and chronic asthma severity.

However, based on our a priori power calculation, we did not meet our enrollment estimates of 800 patients per treatment group. Both lower-than-expected sample size and less-frequent return utilization than our a priori estimates may mean that small differences in outcomes by treatment could have been missed by our analyses. Post hoc analysis by modeling our treatment weights for a larger sample size suggests that differences in return utilization would remain nonsignificant (Supplemental Information). In addition, if the effect size we demonstrated is correct (60% increased odds of reutilization after treatment with dexamethasone), this may be considered a nominal clinical difference given the rarity of this outcome (baseline reutilization rate of only 2.2% for the prednisone). Our confidence intervals around this main effect (OR 1.61; 95% CI, 0.80–3.31), indicates that the actual effect of dexamethasone could be better or worse than prednisone. However, any real small increases in reutilization from dexamethasone that we were underpowered to detect may not offset some of the other advantages of dexamethasone.

Dexamethasone’s favorable side effect profile and pharmacokinetics compared with prednisone may mitigate poor adherence to longer steroid courses, which likely contributes to inadequate symptom resolution and unplanned healthcare reutilization. Administering a second dose of dexamethasone 18 to 24 hours after initial administration allows completion of the full course before hospital discharge. This is notable as data suggest postdischarge medication adherence is suboptimal; only 55% of children filled a postdischarge prescription for systemic corticosteroids, and only 64% reported administering all prescribed doses.^18,19  Beyond medication adherence, dexamethasone reduces postdischarge prescriptions and simplifies medication plans that also likely include 1 or more inhaled medications.

Our analysis has several limitations. First, including a single center limits generalizability. Second, our study captured only return utilization (ED, urgent care, and rehospitalization) in our system, and excluded all primary care visits. In our healthcare system, 90% of children living in the local county who require hospitalization for any cause were admitted to our hospital, and 65% of children seeking emergency care were seen within our system.²⁰  In addition, misclassification for return utilization because of seeking care at a nonaffiliated facility should be independent of steroid treatment choice, and, therefore, should not bias the return utilization outcome on the basis of steroid selection. Third, although we controlled for some confounders with propensity score methodology, our analysis is subject to unknown confounders that may influence the treatment decision and outcomes; this limitation could only be addressed by conducting a large randomized controlled trial. Fourth, although our propensity score model included a categorical variable for receipt of prehospital steroids to account for confounding on the provider decision of hospital administered systemic steroids, our modeling did not include prehospital duration of treatment. Inclusion of patients who received prehospital administration may result in biasing our findings toward no difference, given that the majority of prehospital steroids were prednisone. However, our focus was on hospital-based care and medications in the purview of pediatric hospitalists.

Our study did not identify a difference in 30-day return utilization, length of stay, or need for PICU care by treatment, dexamethasone versus prednisone, for children hospitalized with an asthma exacerbation. Dexamethasone treatment of children admitted for asthma exacerbation appears to be safe and effective.