Clinical Outcome and Antibiotic Dosing Differences by Weight in Children With Acute Osteomyelitis

This retrospective cohort study included children hospitalized with AHO identified in the Cerner Health Facts (HF) database. The HF database is a large clinical resource that includes deidentified, Health Insurance Portability and Accountability Act–compliant data from 664 health care facilities across the United States representing ∼69 million unique patients (adults and children) and >500 million encounters over the past 2 decades. HF data are maintained by the Cerner Corporation (Kansas City, MO). HF contains detailed patient-level clinical data for both inpatient and outpatient encounters at participating facilities, including the following data most relevant to this study: demographic information (eg, patient race and/or ethnicity, all insurance types and self-pay), anthropometric measures, and inpatient medication orders. All work with HF has been deemed nonhuman subjects research by the institutional review board at our organization.

All children aged 2 to 17 years and hospitalized with a primary AHO discharge diagnosis code between January 1, 2010, and December 31, 2017, were included. Diagnosis codes were based on a previously validated process by using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for acute osteomyelitis (730.00–730.08) and unspecified osteomyelitis (730.20–730.28).^6,7  For children hospitalized after September 2015, International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes for AHO (M86.00–M86.08) and other acute osteomyelitis (M86.10–M86.18) were used. All dates within the HF database are shifted for data deidentification purposes.

Children with concurrent ICD-9-CM or ICD-10-CM diagnostic codes indicating chronic comorbid conditions (eg, immunodeficiencies, malignancies, etc) that could complicate the clinical course were excluded, as were those with subacute, chronic, or multifocal osteomyelitis (Supplemental Table 4).^6,7  Patients with ICD-9-CM, ICD-10-CM, and/or Current Procedural Terminology (CPT) codes representing postoperative cases of osteomyelitis were also excluded (Supplemental Table 5). Additionally, if complete anthropometric data were not available during the initial AHO hospitalization encounter, height data were imputed from all encounters occurring within 28 days before or after the initial hospitalization encounter. Patients were excluded if complete anthropometric data could not be identified. Those <2 years of age were excluded because of the lack of an agreed-on definition for obesity in this age group. Children not receiving any antibiotics were excluded because any child being treated for AHO would receive at least 1 antibiotic during their hospital stay. Lastly, patients with extreme LOSs for AHO (<2 or >14 days)^6,7  were excluded (Fig 1).

The primary study exposure was weight category in patients with AHO. Patients were categorized into 1 of 3 weight categories by using measured height and weight data to calculate BMI: healthy weight, overweight, and obesity. These categories were based on the current Centers for Disease Control and Prevention classification system for children, which uses BMI percentile for age and sex to categorize patients into the following weight categories: healthy weight (BMI fifth to 84th percentile for age and sex), overweight (BMI 85th–94th percentile for age and sex), and obesity (BMI ≥95th percentile for age and sex).^10,24  Obesity severity was further defined among children with obesity: class I obesity (BMI 95%–<120% of the 95th BMI percentile for age and sex), class II obesity (BMI 120%–<140% of the 95th BMI percentile for age and sex), and class III obesity (BMI ≥140% of the 95th BMI percentile for age and sex). Children who were underweight were excluded for the purposes of this analysis.

The primary study outcomes included (1) the proportion of children undergoing AHO-related procedures and (2) the proportion of children experiencing AHO-related complications either during or up to 6 months after the index hospitalization. Procedures examined included any AHO-related surgical intervention during the index hospitalization (eg, bone abscess drainage, joint arthroscopy for drainage, etc) (Supplemental Table 6). Complications of AHO included particular diagnoses during the index hospitalization (ie, pathologic fracture, pyogenic arthritis, sepsis). As secondary outcomes, we assessed differences in LOS and treatment failure across weight categories. Treatment failure was defined as any hospital or emergency department (ED) revisits related to AHO (ie, possessing ICD-9-CM and ICD-10-CM codes for AHO listed above) occurring within the 6 months after the index hospital encounter (eg, revisits to the ED, hospital readmissions for AHO) (Supplemental Table 6).⁷ 

Lastly, we assessed differences in mean weight-adjusted daily dose of the most frequently ordered intravenous antibiotics during hospitalization. Drugs included in the analysis were the 4 most commonly ordered antibiotics during AHO hospitalization: clindamycin, ceftriaxone, cefazolin, and vancomycin. We selected the first order for each of these drugs with complete dosing information (ie, drug dose and frequency). Data collected included the drug prescribed, prescribed dose, frequency, route of administration, and formulation. Only intravenous (IV) orders were included in the analysis because initial IV therapy is standard of care for AHO.

For our secondary outcome analysis of antibiotic dosing by weight category, we used total daily dosing of commonly prescribed IV antibiotics to compare differences in dosing between weight categories. Daily dosing for each drug was calculated by using the mean weight-adjusted total daily dose per kilogram of total body weight. The total daily dose for each antibiotic was calculated by multiplying the recorded milligrams per dose by the number of doses prescribed for the first 24-hour period and was recorded as milligrams per kilogram per day.

To provide context for antibiotic doses among children with weights high enough to require the maximum recommended daily doses, we also examined the proportion of the 4 most commonly prescribed drugs that were at or above a maximum daily dose recommended by the Infectious Disease Society of America or as found in Lexicomp.²⁵  These maximum daily doses were determined for each drug by selecting the highest of either the child or adult daily maximum recommended dose for treatment of acute osteomyelitis or severe methicillin-resistant S aureus infections. These maximum daily doses were 4000 mg daily for ceftriaxone, 12 000 mg daily for cefazolin, 4800 mg for clindamycin, and 3600 mg for vancomycin.

For the analysis, children were assigned to 3 possible weight categories (healthy weight, overweight, and obesity). Demographic characteristics were compared across weight categories, with Pearson’s χ² test and analysis of variance tests for categorical and continuous variables, respectively. The Pearson’s χ² test was used to assess statistical associations between weight category and (1) treatment failure and (2) associated conditions and/or surgical interventions. The Kruskal-Wallis test was used to assess statistical associations between weight category and (1) LOS and (2) initial median total daily dose of each antibiotic received during the index hospitalization. All statistical analyses were performed by using SAS software, version 9.4 (SAS Institute, Inc, Cary, NC).

A total of 755 children were included in the analysis (Table 1). Included children were on average 9.4 years of age, 499 (66%) were male, 448 (59.3%) were white, and 336 (44.5%) had government insurance. A majority had a healthy weight (455, 60.3%), whereas 117 were overweight (15.5%), and 183 (24.2%) had obesity. Of children with obesity, 114 (62.3%) had class I obesity, 49 (26.8%) had class II obesity, and 20 (10.9%) had class III obesity.

Of 755 children, 99 (13.1%) underwent an AHO-related procedure during the index and subsequent hospitalizations. The majority of procedures (n = 94; 94.9%) occurred during the index hospitalization. Children with overweight and obesity were more likely than children of a healthy weight to undergo a procedure (18.8% and 16.9% vs 10.1%, respectively; P = .009). Results were similar when examining procedure rates during the index hospitalization encounter (Table 2). No differences existed between patients with class I obesity and class II or III (severe) obesity (P = .38). The most frequent procedures were similar across all weight categories, with the top 2 procedures in every weight category being surgical debridement and drainage of bone abscesses (Table 3).

Complications (pyogenic arthritis, myositis, sepsis) occurred frequently among the cohort, with 273 (36.2%) children experiencing an AHO-related complication during either the index encounter or subsequent encounters. The majority of these occurred during the index hospitalization (n = 214; 78.4%). There were no differences in rates of complications between weight categories (Table 2). Additionally, there were no differences between patients with class I obesity and class II or III (severe) obesity (P = .27).

The overall median LOS for all children was 5.1 days (interquartile range 3.7–7.5). Children with overweight and obesity had longer LOS compared with children with a healthy weight (5.7 and 5.8 vs 4.9 days, respectively; P = .03) (Table 2). Patients who underwent a surgical procedure during their encounter were more likely to have a longer LOS compared with children who did not require a procedure (6.3 vs 5.0 days; P = .0045). Of the 755 included children, 83 (11.0%) had a hospital revisit (ED visit or hospital readmission) within 6 months of the index hospitalization. The prevalence of readmission did not vary between weight category (P = .133), or between patients with class I and class II or III obesity (P = .17).

The most common first antibiotics with complete dosing information prescribed among our cohort included clindamycin (n = 330; 50.3%), vancomycin (n = 113; 17.2%), cefazolin (n = 83; 12.6%), and ceftriaxone (n = 26; 3.9%). All other antibiotics represented <16% of all initial medication orders, with no single antibiotic representing >2% of all antibiotics. The proportion of the 4 most commonly prescribed antibiotics did not vary significantly by weight category (P = .388). When examining differences in daily doses by weight category, we found that doses of the common antibiotics varied for cefazolin (P < .05), clindamycin (P < .05), and ceftriaxone (P ≤ .01), with children with obesity having significantly lower adjusted total daily doses when compared with children with a healthy weight (Fig 2).

When examining the proportion of doses that reached the daily maximum recommended dose for each of the 4 most commonly prescribed drugs, we found that few reached or exceeded the daily maximum recommended dose. Of 534 prescriptions, only 17 (3.2%) doses reached or exceeded the recommended daily maximum dose: 13 of 111 total vancomycin prescriptions and 4 of 25 total ceftriaxone doses. Of these 17 above-maximum doses, 8 (47%) occurred in healthy weight children, 3 (18%) in children with overweight, and 6 (35%) in children with obesity. No clindamycin or cefazolin doses reached the daily recommended maximum dose.

In this study, using a nationwide clinical database to examine clinical outcomes and drug dosing variability among children hospitalized with AHO of varying weight categories, we found that children with overweight and obesity were more likely than children with a healthy weight to undergo procedures during both the index and subsequent hospital encounters. Although there was no difference in rates of complications or treatment failure experienced, children with overweight and obesity were also found to have longer LOS compared with children with a healthy weight. Among the most commonly prescribed antimicrobial agents for AHO, children with obesity were more likely to receive lower weight-adjusted doses overall compared with children with a healthy weight.

Our findings add to previous work examining outcomes in children with obesity who are hospitalized for various illnesses, illustrating that children with obesity may be more likely to experience negative clinical outcomes or complications during their illness course.^16–18  Obesity is known to negatively affect bony architecture in children^19,26  and is associated with elevated baseline inflammation¹¹  and impaired cellular immunity,²⁷  all of which may lead to increased risk of more severe and/or invasive AHO infection. These inherent physiologic risks, coupled with our findings suggesting increased risk of surgical intervention among children with overweight and obesity, draw attention to the fact that this population of children requires specialized awareness or adjustment of protocols to account for this increased risk immediately after AHO diagnosis.

In contrast to previous literature identifying worse outcomes for children with obesity hospitalized for other problems (eg, critical illness, oncologic diagnoses, hospital adverse events),^16,18  we found that children with overweight and obesity hospitalized for AHO experienced no difference in complication rates (eg, sepsis, septic arthritis), ED revisits, or readmissions compared with children with a healthy weight. Children with overweight and obesity in our cohort did have longer LOS, which was associated with undergoing surgical procedures during their stay. To our knowledge, no previous pediatric studies have specifically examined the relationship between obesity and AHO outcomes. In previous studies, researchers have found that certain factors may be associated with increased risk of AHO complications for children in general, including having elevated inflammatory marker levels on presentation and infection with certain bacteria (eg, methicillin-resistant S aureus).^2,9  Further research is necessary to determine the degree to which these associations may exist for children with obesity.

Obesity-specific drug dosing guidelines for children are lacking for the majority of antimicrobial agents, including the drugs included in this analysis.²³  Obesity-specific guidance from the Infectious Disease Society of America does exist for vancomycin prescribing for adults, as well as for children with severe methicillin-resistant S aureus infections (which may apply to some children with AHO),²⁵  but guidelines for children with obesity do not exist for the vast majority of drugs or conditions. This lack of guidance may lead to variability in dosing for children with obesity.^22,28  Our findings reveal that children with overweight and obesity were more likely to receive smaller weight-adjusted daily doses of their antibiotics, which may put them at risk for treatment failure. Previous work examining prescribing patterns for clindamycin revealed reductions in prescribing variability when pharmacokinetic data were available for children with obesity.²⁹  Ideally, the creation of evidence- and expert opinion–based dosing guidelines for a wide array of drugs would help reduce variability in prescribing practices for children with obesity.

Importantly, the optimal dosing strategy for the included antibiotics among children with overweight and obesity and AHO is unknown because of a lack of pharmacokinetic data in this population. Evidence in this area has been growing in recent years, however. A recent population pharmacokinetic analysis suggested that total body weight dosing of clindamycin should provide adequate bone concentrations for the treatment of AHO among children with obesity.²⁹  This study also suggested that doses exceeding the recommended clindamycin adult dose were likely unnecessary. However, less is known about the appropriate cefazolin dosing among children with overweight and obesity and AHO. In a small prospective pharmacokinetic study, Koshida et al³⁰  reported similar volume of distribution and clearance between children with and without obesity. Conversely, another study identified alterations in cefazolin volume of distribution among adults with obesity compared with adults of a healthy weight,³¹  and two additional adult pharmacokinetic studies also revealed decreased cefazolin subcutaneous tissue distribution among various patient populations with obesity.^32,33  Vancomycin dosing is another drug with a growing evidence base of pharmacokinetic data in patients (mostly adults) with obesity; current evidence reveals that loading doses are modulated by volume of distribution, whereas maintenance dosing is modulated by drug clearance. Both volume of distribution and drug clearance are altered in patients with obesity, so early and frequent therapeutic drug level monitoring is recommended.²⁵  In our study, we found no significant differences in complication rates or treatment failure (ie, ED revisits or readmissions within 6 months of the index encounter) in children with obesity, suggesting treatment strategies (both inpatient and subsequent outpatient or oral therapies) were adequate despite overall lower antibiotic dosing and a paucity of pharmacokinetic data to support evidence-based dosing recommendations. However, further investigation is needed to better understand antibiotic exposure in children with obesity. With clear pediatric pharmacokinetic data lacking, particularly for children with obesity,^21,23  the ability for providers to make optimal antibiotic dosing decisions remains difficult.

This study should be viewed in light of some limitations. Although the HF database provides a wealth of patient-level information not available in many other large data sets, the data collected may be incomplete or inconsistent across contributing organizations.³⁴  For example, it is possible inaccurate or incomplete primary discharge diagnosis codes may have misidentified or failed to include all relevant cases of AHO hospitalization. Additionally, when examining 6-month revisit rates, the HF database only captures encounters that occur at each participating institution; therefore, it is possible some revisits or readmissions were missed if patients sought care at another facility. Anthropometric data are also often incompletely recorded during hospitalizations³⁵ ; however, our study cohort included a substantial proportion of patients with overweight and obesity. Importantly, our analysis also lacked more detailed patient-level information (eg, laboratory and microbiology results, illness severity, central line use, radiologic data) that would strengthen the overall results and allow us to control of confounding factors. This type of clinical data should be considered in any future studies. Additionally, we were unable to account for differences in outcomes for children with more severe obesity because of limitations in sample size for class II and III obesity. In our analysis of antibiotic dosing, we included only the first medication order for each included antibiotic. Although this reflects the intent of initial therapy, it is not indicative of the antibiotic choice or dosing for the entire course of therapy for AHO, which should be considered when examining longer-term outcomes and complications of AHO.

Children with overweight and obesity are more likely to undergo procedures and have longer LOS during hospitalization for AHO compared with children with a healthy weight. Children in higher weight categories also receive lower weight-based dosing of antibiotics, aligning with previous work indicating variability in drug dosing for children with obesity. Special considerations should be made when caring for children with overweight and obesity who are hospitalized for treatment of AHO to mitigate these disparate outcomes. Additionally, future research and improvement interventions should focus on standardization of care and creation of antibiotic dosing recommendations that include children with overweight and obesity.