Comparative Effectiveness of Dexamethasone Versus Prednisone in Children Hospitalized With Acute Croup

We conducted a retrospective, multicentered cohort study using the Pediatric Health Information System database, a comparative administrative and billing database, including clinical and resource utilization data for inpatient, ambulatory surgery, emergency department, and observation encounters from 49 tertiary care children’s hospitals, which accounts for approximately 20% of all pediatric hospitalizations in the United States.¹⁵  Deidentified data are submitted by participating hospitals. An encrypted medical record number permits identification of readmissions and resource utilization at the same hospital. In accordance with the policies of the University of California San Diego’s institutional review board, this research, using a deidentified data set without any protected health information, was not considered human subjects research, and exemption status was obtained.

Children between the age of 6 months to <6 years were eligible for inclusion when hospitalized between January 1, 2015 to December 31, 2019 with an International Classification of Diseases (ICD), 9^th and 10^th Revision, principal diagnosis of acute croup (ICD-9: 464.4 and ICD-10: J05.0).

The age range was chosen because croup typically affects children between 6 months to 6 years of life,¹⁶  and as the child’s airway enlarges with growth, the incidence declines rapidly.¹⁷  In the Western Australian study, the mean age of the cohort was around 30 months with a SD of 17 months,⁸  and other previous randomized control trials used upper age limit between 6 and 10 years.^6,7  To identify a population of typical croup patients requiring hospitalization, we excluded children who required direct ICU admission, those with significant underlying comorbidity in the form of complex chronic conditions¹⁸  or who were transferred in from an outside hospital where previous treatment may have been administered as well as those whose length of stay exceeded 5 days which accounted for 0.5% of the study cohort to minimize the effect of extreme outliers. Hospitals without LOS reported in hours were also excluded.

Other concurrent secondary airway obstruction diagnoses where corticosteroid may have been administered were excluded: acute epiglottitis (ICD-9: 464.3, 464.31; ICD-10: J05.10, J05.11), supraglottitis (ICD-9: 464.5, 464.51; ICD-10: J04.30, J04.31), laryngitis (ICD-9: 465.0, 464.2; ICD-10: J06.0, J04.2), tracheitis (ICD-9: 464.1, 464.11; ICD-10: J04.10, J04.11), and asthma (ICD-9: 493; ICD-10: J45) (Fig 1).

Corticosteroid treatment of acute croup was classified as dexamethasone for children who received dexamethasone only, and as prednisone or prednisolone for children who received either prednisone or prednisolone only. Patients who received a combination of both or neither medication were excluded from the analysis. Both parenteral and enteral formulations of the medications were included.

The main outcome measure was total LOS for the index hospitalization calculated in hours on the basis of admission and discharge time. LOS was expressed as geometric mean, which is less sensitive to outlier effect because of the skewed nature of the data. Secondary outcome measures included escalation of care to the ICU after admission, administration of heliox (helium-oxygen mixture), requirement of mechanical ventilation or noninvasive positive pressure ventilation, or any surgical intervention in the forms of laryngoscopy and bronchoscopy during the index encounter. Additional outcomes included 7-day rate of all-cause readmission and return to the ED.

Model covariates included patient demographics including age, sex, race, payor, diagnostic tests received, such as chest radiograph and neck radiograph,¹⁹  and treatment received, such as antibiotics, nebulized racemic epinephrine, inhaled steroid and β agonist. Severity was measured using hospitalization resource intensity score for kids,²⁰  which is a relative cost weight based on each discharge’s assigned All Patient Refined Diagnosis Related Group and level of severity (3M Health Care).

Characteristics of the exposure groups were summarized using frequencies and percentage for categorical variables and compared by using χ² tests. To address the possibility of confounding by indication, we applied a 1:1 propensity score-matching method. Propensity score were developed using a multivariable logistic regression model with corticosteroid type as the dependent variable and age, payor, receipt of racemic epinephrin, and hospitalization resource intensity score for kids as fixed effects. Dexamethasone and prednisone or prednisolone recipients were matched on propensity score using a greedy nearest-neighbor matching with a caliper set at one-quarter of the SD of the logit of the propensity scores. We forced matches within hospitals to avoid across-hospital variability biases. Outcomes in the matched cohort were compared by using generalized estimating equations with binomial or normal distributions as appropriate and controlling for hospital clustering. Continuous variables were log-transformed before modeling. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) with P < .05 being considered statistically significant.

There were 11 740 hospital encounters for acute croup across 39 children’s hospitals that met the inclusion criteria during the study period. Dexamethasone was administered in the majority of the patients 11 260 (95.9%). Thirty five of the 39 hospitals used prednisone or prednisolone, and the use was quite variable, ranging from 0.03% to 36.4% with a median of 1.33%. Children receiving prednisone or prednisolone were more likely to be in the 1- to 3-year age group and more likely to have private insurance. The older children between the age of 4 to 5 made up a small portion of the entire cohort (11.3%), and a small portion of each of the treatment groups with 11.3% in the dexamethasone group and 12.3% in the prednisone or prednisolone group (Table 1). They were also less likely to receive racemic epinephrine, and with slightly lower severity score compared to the dexamethasone-treated children. Although the use of chest radiograph was similar in each group, the children who received prednisone or prednisolone were more likely to have a neck radiograph during the hospital encounter (Table 1).

Propensity matching resulted in a study population of 960 patients, with 480 patients receiving dexamethasone and 480 patients receiving prednisone or prednisolone in 35 hospitals. We were able to balance all covariates between the groups (Table 2). In the matched cohort, the geometric mean LOS in hours were not significantly different between the dexamethasone group (20.3±1.9) and prednisone or prednisolone group (18.5±1.8) (P = .35). Patients who received dexamethasone were more likely to require ICU transfer with an adjusted odds ratio of 21.92 (P = .01), and all other analyzed outcomes were equal between the 2 groups with no significant differences in mechanical ventilation, heliox use, or surgical interventions. The overall incidence of ICU transfer was low with 4.4% in the dexamethasone group and 0.2% in the prednisone or prednisolone group. The 2 groups also had very similar 7 day return rates to the ED (1.3%–2.3%, P = .24) as well as 7-day readmission rates (2.1% to 3.1%, P = .37) (Table 3), which were not statistically significant.

In addition, we performed a subgroup analysis on those children who received racemic epinephrine during their hospital encounter, which may be an indicator for higher acuity. The mean LOS were not significantly different between dexamethasone group (23.8±1.9) and prednisone or prednisolone group (22±1.8) (P = .36). Although there were no ICU transfers in the prednisone group, all other analyzed outcomes remained equal between the 2 treatment groups and the 7-day return rates to the ED (1.2%–2.3%, P = .37), and readmission rates (3.1% to 4.3%, P = .51) remained low and not significantly different between the groups (Table 3).

In this multicenter retrospective cohort study, we found that children who received dexamethasone or prednisone or prednisolone both had very short LOS with very few patients requiring respiratory support such as mechanical ventilation, noninvasive positive pressure ventilation, or heliox. The overall rate of surgical intervention is low at 0.8% to 2.9% with no difference between the 2 treatment groups. Although the rate of ICU transfer was higher in the dexamethasone group when compared to the prednisone or prednisolone group, the overall rates were very low.

Previous reporting using propensity score matching for acute asthma hospitalization has shown effectiveness of both dexamethasone and prednisone or prednisolone with the prednisone or prednisolone as the traditional standard of care.¹³  It is interesting to note that the percentage of administration is completely reversed with dexamethasone being the predominant medication given in acute croup, whereas prednisone or prednisolone is given in majority of the patients hospitalized for acute asthma although there are many similarities in the presentation of these 2 common acute respiratory illnesses with croup affecting the larger airway and asthma smaller airways. The reason for this practice difference is unclear. It is possibly that the respiratory distress is more acute in presentation for croup compared to asthma that dexamethasone is preferred for its faster onset of action and longer half-life, which requires less frequent dosing.²¹  Given that the targeted action of glucocorticoids in both acute croup and asthma is to reduce airway inflammation,⁵  prednisone or prednisolone may be considered as reasonable treatment alternatives to dexamethasone in acute croup patients.

In the overall population studied here, 86.6% of the patients received racemic epinephrine, and 53% of those receiving prednisone were treated with racemic epinephrine during the hospital encounter. Although often used in the management of acute croup, in previous randomized controlled trials in ED population, a much lower rate of racemic epinephrine was reported. In fact, in the Australian ED study, racemic epinephrine was only given in 2.2% to 3% of the patients.⁸  The higher rate of racemic epinephrine administration in this hospitalized cohort further demonstrated the higher acuity of the presentation in hospitalized patients. The subgroup analysis on only those children who received racemic epinephrine (Table 3) showed the similar outcome between dexamethasone and prednisone or prednisolone with short LOS and a low rate of ED return or readmissions.

There are several limitations to the study. It is an observational study and causality cannot be concluded on the basis of this data. Despite several randomized control trials demonstrating equivalent efficacy of dexamethasone and prednisone or prednisolone in both ambulatory and ED cohorts, this topic has not been formally addressed in the hospitalized children who are likely to present with higher acuity and more severe symptoms of acute croup compared to the ED patients. Given the low incidence of acute croup admission, a prospective double-blinded randomized control study for the comparison of the 2 corticosteroid treatment groups will require a multicentered study with a long recruitment period to reach adequate power. Propensity score matching is a statistical method using observational data that attempts to estimate the effect of a treatment, in our case, the type of corticosteroid. We used propensity score matching to reduce bias in estimating the treatment effects but we may not have considered or matched all patient factors, and some residual bias may still exist. There may be local hospital practice differences that were not accounted for in this study given that the use of prednisone or prednisolone varied from 0% to 36.4%. Lastly, because of the administrative nature of the database, we are unable to report on the exact dosing used in this cohort, and there may be significant variabilities in dosage and frequency used across different centers.

In conclusion, we found that children hospitalized with acute croup have relatively low acuity with an overall short hospital LOS and low rate of ED return or readmission. We did not find any difference in hospital resource utilization among children who were treated with dexamethasone versus prednisone or prednisolone with respect to surgery and respiratory support. Although the dexamethasone group had high rate of ICU transfer, the overall incidence for escalation of care remains low. Prednisone or prednisolone may be considered as a reasonable alternative to dexamethasone for the treatment of children hospitalized with acute croup, giving clinicians broader treatment options. Future prospective randomized controlled trials are necessary to establish truly equivalent efficacy and will likely require multicentered efforts.