Intravenous Dexamethasone Use and Outcomes in Children Hospitalized With Septic Arthritis

Septic arthritis, a serious synovial joint infection, commonly occurs in children, with an estimated incidence of 5 to 12 cases per 100 000 children per year.¹ Septic arthritis typically results from hematogenous spread following disruption of the skin barrier and predominantly involves large joints (ie, hip or knee). Children aged younger than 3 years are at the highest risk for these infections.² As the acute phase response elicited by invading bacteria can begin to cause irreversible joint damage within hours of joint entry,³ prompt diagnosis and treatment can mitigate long-term morbidity.

The standard paradigm for initial treatment of septic arthritis consists of joint debridement and irrigation with administration of intravenous antibiotics. However, 10% to 25% of children develop synovial membrane and cartilage destruction despite early intervention.^4,5 Adjunctive systemic corticosteroids, which attenuate the cytokine-mediated inflammatory response, have been hypothesized to improve outcomes of children with septic arthritis.^4,6,7 Yet current data supporting this practice are highly limited. A 2018 Cochrane review that included 149 children from 2 randomized controlled trials (RCTs) deemed the overall evidence for corticosteroids to be low quality due to attrition bias and selective reporting.¹ Moreover, through immunosuppression, corticosteroids have the potential to exacerbate bacterial infections.⁸ Notably, the 2023 Clinical Practice Guideline on Diagnosis and Management of Acute Bacterial Arthritis developed by the Pediatric Infectious Disease Society and Infectious Disease Society of America recommends against using adjunctive corticosteroids, noting that this recommendation is conditional due to “very low certainty” of evidence, which “places a high value on avoiding potential serious harms despite providing potential minimal beneficial effects.”⁹ Meanwhile, a 2018 position statement from the Canadian Paediatric Society outlining management of acute osteoarticular infections in children omits discussion of the use of steroids.¹⁰ 

Given the paucity of evidence, a multicenter administrative data set provides a unique opportunity to study the comparative effectiveness of systemic corticosteroids in this population. We aimed to (1) describe variation in systemic corticosteroid use in children with septic arthritis and (2) assess its association with clinical and utilization outcomes in a large retrospective cohort of children hospitalized with septic arthritis.

We performed a multicenter retrospective cohort study using the Pediatric Health Information System (PHIS; Children’s Hospital Association [CHA], Lenexa, KS) database, which contains administrative and billing data from 48 children’s hospitals across the United States. Patients are deidentified before inclusion in the database; however, encrypted medical record numbers allow for tracking of individual patients across hospitalizations within a hospital. Available data include demographics, discharge diagnoses, procedure codes, and date-specific laboratory, imaging, and medication data. Data quality and reliability are assured through a joint effort by CHA and participating hospitals. This study, for which we used a deidentified data set, was not considered human subjects research in accordance with the policies of the senior author’s local institutional review board.

We included hospitalized children and youth aged 2 months to 21 years (referred to as “children” throughout for simplification) who were discharged from PHIS-participating hospitals with a diagnosis of septic arthritis between January 1, 2016, and December 31, 2020, following either an inpatient or observation status stay. Septic arthritis was defined as the presence of an International Classification of Diseases, 10th Revision, Clinical Modification discharge diagnosis code for septic arthritis (M00.x) as either a principal or secondary diagnosis, consistent with prior PHIS studies.^11,12 We excluded patients for the following reasons: (1) hospitalization in the prior 30 days, which suggests the possibility of a hospital-acquired condition; (2) no antibiotics administered, which makes bacterial arthritis unlikely; (3) alternate diagnoses that would predispose to severe infection or warrant different management (eg, primary immune deficiency, sickle cell disease, malignancy, trauma, and nonclassic infection; Supplemental Table 1); and (4) no joint drainage within the first 2 days of hospitalization, which would be inconsistent with the standard of care. Joint drainage procedures were identified using International Classification of Diseases, 10th Revision, Procedure Coding System codes and included procedures performed in the operating room as well as other locations (eg, bedside procedures in the emergency department [ED]). Hospitalizations were excluded if the length of stay (LOS) exceeded 28 days (99th percentile of hospitalizations) or if LOS data in hours were not available.

The primary exposure was receipt of at least 1 dose of intravenous dexamethasone within the first 2 days (days 0–2) of hospitalization. We selected this exposure window to limit the treatment group to children for whom dexamethasone was used early in their clinical course and to permit association with short-term inpatient outcomes. Children receiving an initial dose of intravenous dexamethasone after day 2 of hospitalization, dexamethasone via other route, or an alternative corticosteroid were classified as not receiving adjunctive dexamethasone.

In identifying patient-level associations between intravenous dexamethasone use and clinical outcomes, our primary outcome of interest was hospital LOS (measured in hours). Hospitalization costs (estimated from charges using hospital- and year-specific cost-to-charge ratios) were measured as a secondary outcome. To assess hospital utilization, we also measured rates of postdrainage computed tomography (CT) or magnetic resonance imaging (MRI) (defined as occurring on, or after, the day following procedural drainage), opioid analgesic use, and repeat drainage procedures. Finally, we measured rates of acute care reutilization (ED visit or rehospitalization) within 30 days of discharge.

We included age, sex, race and ethnicity (non-Hispanic white, non-Hispanic Black, Hispanic, Asian, and Other) and payor as demographic characteristics. Race and ethnicity was included as a variable to assess for potential differences in treatment that may reflect systemic racism and/or bias. We measured the presence of complex chronic conditions¹³ to capture clinical complexity. To account for additional differences in clinical course and/or management that may be associated with both intravenous dexamethasone use and our outcome measures, we measured transfer in from another institution, operating room charges (to indicate surgical procedures specifically occurring in the operating room), antibiotic exposure, and use of washout antibiotics. Antibiotic exposure was defined as occurring in day 0 or 1 of the hospitalization and was categorized based on antibiotic monotherapy or combinations that would be anticipated to provide varying degrees of coverage for causative organisms, including methicillin-resistant Staphylococcus aureus (MRSA) and Kingella species. Washout antibiotics were defined as those used for intraoperative joint irrigation. Because adjacent osteomyelitis may influence management and outcomes, we also identified this condition using International Classification of Diseases, Tenth Revision (ICD-10) codes previously established for PHIS database research (primary or secondary diagnosis codes for M86.00-M86.09, M86.10-M86.19, M86.20-M86.29, M86.8X0-M86.8X9, A02.24, or M86.9).¹⁴ For descriptive purposes, we assessed for causative pathogens using organism-specific ICD-10 discharge diagnosis codes and measured other corticosteroid routes and agents used during hospital days 0 to 2.

We described characteristics of the exposure groups using frequencies and percentages and compared exposure groups using χ² tests. We also examined dexamethasone use at the hospital level, describing the percentage of patients receiving dexamethasone at each hospital.

In our primary analysis evaluating the association between dexamethasone use and outcomes, to account for systematic differences between patients who did and did not receive dexamethasone, we calculated propensity scores for each patient using the following covariates that were hypothesized to confound the association between dexamethasone use and outcomes: sex, race and ethnicity, transfer in from another institution, operating room use, presence of a complex chronic condition, antibiotic category, and use of washout antibiotics. We matched patients 1:1 on the nearest-neighbor propensity score using a greedy algorithm and forced matches on age group and hospital. We used generalized estimating equations to account for hospital-level clustering for categorical outcomes (post-drainage CT/MRI imaging, opioid use, repeat drainage procedures, and 30-day reutilization) and continuous outcomes (LOS and cost) after log transformation. We performed a subanalysis excluding patients with osteomyelitis to evaluate for any differences in outcomes in this population.

Analyses were performed using SAS version 9.4 (SAS Institute, Inc), and statistical significance was defined as P < .05.

We identified 3524 hospitalizations for septic arthritis across 47 hospitals that met inclusion criteria (Figure 1). All hospitalizations from one of the 48 PHIS hospitals were excluded because LOS was not reported in hours. Most hospitalizations were for children aged 1 to 9 years (67.5%), male children (58.2%), and children who were non-Hispanic white (57.8%) (Table 1). More than 90% of children had a code indicating operating room charges (exclusive of bedside procedures), whereas less than 5% of children required intensive care unit (ICU) admission. Antibiotic regimens varied substantially across hospitalized children; cephalosporins alone or in combination with clindamycin or vancomycin were prescribed most frequently (70.9%). Nearly 58% of hospitalizations did not identify a specific causative organism within billing codes (Supplemental Table 2); however, 25.9% of hospitalizations included a diagnosis code for methicillin-susceptible Staphylococcus aureus and 7.6% a diagnosis code for MRSA. Corticosteroid administration using an agent or route other than intravenous dexamethasone during days 0 to 2 was infrequent (1.1% for oral dexamethasone, <1% for other agents). A total of 30.9% of children were also assigned a discharge diagnosis code for osteomyelitis. Demographics and clinical characteristics of the propensity-matched cohort are shown in Supplemental Table 3.

Intravenous dexamethasone was initiated for 1109 children (31.5% of total) with septic arthritis on hospital day 0 to 2. Of those children, 557 (50.2%) received dexamethasone on hospital day 0 and 448 (40.4%) received dexamethasone on day 1. In addition, 995 children (89.7%) received dexamethasone for 1 day and 114 (10.3%) received it for 2 or more days (Supplemental Table 4). The majority of children (965, 87.0%) received dexamethasone on the same day as procedural drainage (Supplemental Table 5). At the hospital level, the median rate of dexamethasone use across the 47 hospitals was 28.4% (IQR, 18.9%–43.5%; full range, 0%–56.8%; Figure 2).

Dexamethasone use was more common among older age groups and those who had an operating room procedure (Table 1). Dexamethasone use was less frequent among children who underwent peripherally inserted central catheter line placement or had an ICU stay. We observed small but statistically significant differences in antibiotic regimens administered between dexamethasone groups.

Unadjusted results for the full study cohort are presented in Table 2. Opioid use occurred frequently (>95% of hospitalizations) in the full study cohort. A total of 17.8% of children received a CT or MRI scan following the day of procedural drainage, and 17.6% underwent repeat procedural drainage. Compared with children who did not receive dexamethasone, those who received dexamethasone had shorter mean LOS and greater opioid use (Table 2).

All baseline demographic and clinical characteristics included in the analyses were balanced in the propensity-matched cohort, and adjusted outcomes are shown in Table 3. The findings of decreased LOS (geometric mean, 100.5 [SD, 1.7] vs 114.3 hours [SD, 1.8], P < .001) and greater opioid use (adjusted odds ratio, 3.80; 95% CI, 1.49–9.70; P = .01) among children who received dexamethasone persisted. Additionally, children who received dexamethasone had significantly lower costs (geometric mean, $16 660.5 [SD, 1.8] vs $18 243.3 [SD, 1.9]; P = .01). There were no significant differences between groups in postdrainage CT or MRI, repeat procedural drainage, or 30-day acute care reutilization (hospital readmission, ED revisit, or either) in either the entire cohort or the propensity-matched cohort. A subanalysis excluding children with concomitant osteomyelitis demonstrated consistent findings (Supplemental Table 6).

In this multicenter observational study, intravenous dexamethasone was used to treat nearly one-third of children with septic arthritis, with substantial variation in use across hospitals. Children in our propensity-matched cohort who received dexamethasone had a mean 13.8 hours shorter LOS and $1583 lower cost per admission without increased odds of 30-day acute care reutilization, postdrainage cross-sectional imaging, or repeat drainage procedures.

This study adds important context to the short-term impact of corticosteroid use for children with septic arthritis. To our knowledge, the existing literature describing the use of dexamethasone in septic arthritis is limited to 2 small RCTs, 1 nonrandomized clinical trial, and 1 observational study. Odio et al conducted an RCT of intravenous dexamethasone (0.2 mg/kg/dose every 8 hours for 12 doses) compared with placebo among children in Costa Rica, finding that dexamethasone reduced residual joint dysfunction for up to 12 months after therapy completion and shortened duration of symptoms by a mean 5.47 days.¹⁵ However, outcomes were evaluated in only 100 of the 123 enrolled children, leading to high risk for attrition bias.¹ Harel and colleagues similarly performed an RCT of intravenous dexamethasone (0.15 mg/kg/dose every 6 hours for 16 doses) compared with placebo among 49 children in Israel and demonstrated a shorter duration of fever and local inflammatory signs, elevated acute-phase reactants, and antibiotic use in the dexamethasone group.¹⁶ The authors also provided a narrative statement that patients treated with dexamethasone had a shorter duration of hospitalization yet did not report data to support this conclusion. This study was also limited by an imbalance of covariates between treatment groups, significant (41%) loss to follow-up at 1 year,^1,9,16 and isolation of Staphylococcus aureus in only 3 patients (6% of total cohort) with a causative organism identified in only 18% of the study population. This is in contrast to the 33% of patients in our study with documented billing codes for S aureus infections, which is more consistent with prior literature and is likely more representative of the epidemiology of septic arthritis in the United States.^17,18 Additionally, neither trial directly reported hospital LOS, readmission rates, or costs, which are outcome measures important to patients, families, payors, and health systems. The nonrandomized studies, respectively including 60 and 116 children at single centers, both similarly demonstrated improvement in clinical and laboratory parameters but were limited by generalizability and absence of long-term outcomes.^19,20 

In our study, which includes a larger and more representative sample of children, the modestly decreased LOS observed among children receiving dexamethasone supports the aforementioned findings. Although symptom control could not be directly measured using PHIS data, and the physicians proactively treating with dexamethasone could also be more proactive about expediting hospital discharge, it is possible that earlier hospital discharge may be attributable to improved symptom control. Alternatively, dexamethasone may have contributed to a faster decline in inflammatory markers,²¹ improved appetite and oral intake,²² and/or increased activity or energy levels,²³ thereby allowing discharge criteria to be met earlier. Of note, although we found a statistically significant difference in the receipt of opioids (99.5% for children receiving dexamethasone vs 98.3% for children not receiving dexamethasone), the trivially higher opioid use among children receiving dexamethasone may simply reflect a broader tendency toward intensive medication prescribing among treating teams.

Regardless, the similar rates of 30-day acute care reutilization and repeat drainage procedures between treatment groups suggests that corticosteroids did not contribute to short-term treatment failure. Specifically, this may provide reassurance that corticosteroids did not lead to recrudescent infections upon discontinuation (due to “masking” of fever or inflammatory signs) or other severe adverse events (eg, gastrointestinal bleeding or severe behavioral disturbances).^24,25 Although our study suggests possible benefit of this practice, we were not able to assess long-term outcome data or subsequent outpatient evaluations that may have differed between treatment groups. This limitation, in conjunction with the potential harms associated with corticosteroids and conflicting evidence of their effectiveness in other acute pediatric infectious conditions including orbital cellulitis and retropharyngeal/parapharyngeal abscesses,^26–28 highlights the need to conduct a large, multicenter RCT to more definitively assess efficacy and safety of corticosteroids among children with septic arthritis using patient-centered outcomes.

The imperative for a future large clinical trial is particularly strong given that, as described in other acute pediatric infectious conditions,^27–29 we found hospital-level dexamethasone use to be highly variable (IQR, 18.9%–43.5%). Because corticosteroid use may be more driven by individual and institutional factors rather than evidence, should our study findings be corroborated in a large RCT using patient-centered outcomes, and supported by revised national guidelines, there is ample opportunity to integrate corticosteroid use into practice through large-scale implementation studies, reduce practice variation, and improve the quality of care for a common and severe infection among children.

Strengths of our study include a large sample size across multiple US hospitals, promoting both internal and external validity, as well as rigorous adjustment for potential confounding factors. However, our study has several limitations. Our use of discharge diagnoses to define our cohort may permit misclassification. We may have inadvertently included children with alternate diagnoses and, less likely but still possible, excluded some children with septic arthritis. It is unclear whether such misclassification would be associated with systematic differences in dexamethasone use. Including patients transferred from another hospital may also introduce misclassification bias if some of these patients may have received dexamethasone prior to transfer; however, based on our clinical experience, we believe that administration prior to transfer would be rare. Additionally, we accounted for hospital transfers within the propensity scores such that this variable was balanced in our propensity-matched cohort, which would be expected to mitigate misclassification bias. Confounding by indication, whereby children predisposed to worse outcomes are most likely to receive dexamethasone, might cause us to underestimate the benefit of dexamethasone. To mitigate this possibility, we identified and excluded children with principal and secondary discharge diagnoses codes representing diagnoses that may warrant systemic corticosteroids (eg, sickle cell disease, cancer, and trauma; Supplemental Table 1). We also used propensity score matching to account for measured differences between dexamethasone recipients and nonrecipients. However, there may be unmeasured confounders that contribute to the observed associations between dexamethasone and outcomes. Furthermore, although the use of other corticosteroid agents and routes was rare, our study design did not assess their impact on clinical outcomes. Additionally, the PHIS hospitals included in the study are dedicated children’s hospitals; further study would be needed to assess generalizability to children cared for in nonchildren’s hospitals.^30,31 Finally, we were unable to assess longer-term patient-centered outcomes (eg, joint function and outpatient visit frequency to orthopedic surgeons), which would be important to capture in a randomized trial.

In this large, multicenter study, dexamethasone use varied widely across hospitals. Children receiving dexamethasone had shorter hospital LOS and incurred fewer costs without increased rates of 30-day reutilization. These findings provide additional support for the potential benefit and safety of dexamethasone in children with septic arthritis. A large RCT to definitively determine its efficacy is warranted before dexamethasone use can be recommended.