Vancomycin Use for Pediatric Clostridium difficile Infection Is Increasing and Associated with Specific Patient Characteristics

Abstract

In adults with Clostridium difficile infection (CDI), enteral vancomycin is considered the preferred initial regimen for severe disease; however, patterns of antimicrobial use for children with CDI are unknown. We sought to describe trends in and predictors of vancomycin use for the treatment of children with CDI admitted to tertiary-care children's hospitals in the United States. We used a database of freestanding children's hospitals to identify patients 1 to 18 years old with CDI between January 2006 and June 2011. The first hospitalization with a diagnosis of CDI for each patient was identified, and CDI-directed therapy was assessed. Generalized estimating equations were used to identify predictors of vancomycin receipt, controlling for clustering within hospitals. Vancomycin use has increased significantly (P = 0.005), with substantial variability between hospitals (0 to 16%). In multivariate analyses, vancomycin use was more common in children age 7 to 13 years old (versus children 1 to 2 years old: adjusted odds ratio [AOR] = 1.57; 95% confidence interval [CI] = 1.13 to 2.18), 14 to 18 years old (AOR = 1.40; 95% CI = 1.11 to 1.76), in an ICU (AOR = 1.37; 95% CI = 1.05 to 1.80), or with chronic gastrointestinal conditions (AOR = 2.01; 95% CI = 1.44 to 2.81). Vancomycin use was less common in black (AOR = 0.53; 95% CI = 0.39 to 0.73) and Hispanic (AOR = 0.63; 95% CI = 0.47 to 0.84) patients and in children with malignancies (AOR = 0.57; 95% CI = 0.36 to 0.89). Despite a lack of empirical evidence to suggest superiority, vancomycin use for pediatric CDI is increasing. Furthermore, there is substantial variability in vancomycin use between hospitals. Further studies are needed to explore potential racial and ethnic differences in CDI management and to investigate clinicians' rationale for using vancomycin for initial therapy in selected populations.

INTRODUCTION

Clostridium difficile is the most commonly recognized cause of infectious diarrhea in the health care setting (1). Recent studies have shown that the incidence of C. difficile infection (CDI) is increasing both among hospitalized pediatric patients and among children in the community (2–6). Despite this apparent increase in incidence, rates of serious outcomes, including mortality, colectomy, and length of stay among children with CDI have remained steady (2, 3). These data suggest that CDI severity in children appears to be relatively stable, in contrast to adult CDI epidemiology.

Metronidazole and vancomycin have been the mainstays of CDI management for several decades. Current Infectious Diseases Society of America (IDSA) and Society for Healthcare Epidemiology of America (SHEA) clinical practice guidelines suggest that metronidazole should be considered the drug of choice for the management of initial CDI episodes in most adult patients (1). For adults with severe disease, enteral vancomycin may be associated with a higher rate of clinical cure and is considered the preferred agent (7). However, pediatric data on optimal CDI therapy are scarce and whether existing definitions of severe disease or recommendations for vancomycin use are applicable to children with CDI remains unclear. A recent American Academy of Pediatrics (AAP) policy statement provides similar recommendations for children, but the degree to which current practice patterns reflect existing data and recommendations is unknown (8).

Given the lack of a standardized definition for severe CDI in children and the absence of pediatric data comparing the efficacy or current prescribing of CDI agents, we sought to describe national trends in the use of vancomycin for the initial treatment of hospitalized children with CDI. Identifying the patients receiving this agent is an important first step in understanding the degree to which adult guidelines have been incorporated into pediatric practice, current perceptions about the optimal management of children with CDI, and identifying populations in whom variations in care may be occurring.

MATERIALS AND METHODS

PHIS database.

This retrospective cohort study was conducted using data from the Pediatric Health Information System (PHIS) database, a large administrative database that captures inpatient data from 43 freestanding children's hospitals affiliated with the Children's Hospital Association (CHA; Overland Park, KS) across the United States. The data quality and reliability are assured though a joint effort between CHA and participating hospitals. For the purposes of external benchmarking, participating hospitals provide discharge/encounter data including demographics, diagnoses, and procedures. The data are deidentified at the time of submission and are subjected to a number of reliability and validity checks before being included in the database (9). For the present study, the 42 hospitals that also submitted resource utilization data (e.g., pharmaceuticals, imaging, and laboratory) were included. The study protocol was reviewed and approved by CHA and the Boston Children's Hospital Committee on Clinical Investigation.

Study population.

We included patients 1 to 18 years of age with discharge dates between 1 January 2006 and 30 June 2011. Patients under the age of 1 year were excluded since infants are often colonized with C. difficile, and a causal relationship between colonization and diarrheal illness has not been established (10–13). Children were included if they had an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) discharge diagnosis of intestinal infection due to Clostridium difficile (008.45). Each hospital provides up to 40 ICD-9-CM diagnosis codes per admission. In an attempt to accurately capture the time of infection, charge data were reviewed, and patients were included only if a diagnostic test billing code for C. difficile was identified. Diagnostic methods captured by the database include C. difficile culture, toxin/toxin gene detection (enzyme immunoassay, PCR), and antigen testing (glutamate dehydrogenase). To analyze CDI management, patients were included if pharmacy charge data indicated receipt of a C. difficile-active agent (oral metronidazole, parenteral metronidazole, or oral vancomycin) 1 day before through 2 days after the date of testing. This strategy has been used in prior PHIS studies and been shown to accurately and reliably identify children hospitalized with CDI (2, 14).

Only the first hospitalization with a discharge diagnosis of CDI was analyzed, and only the first occurrence of CDI testing during the hospitalization was assessed. Patients with a discharge diagnosis of CDI from a hospitalization within the 6 months before the study period were excluded. Using this strategy, we sought to identify the initial therapeutic choice for the management of hospitalized patients with a first episode of CDI.

Primary outcome and independent variables.

The primary outcome for the study was the receipt of at least one dose of vancomycin around the time of CDI testing. Potential predictors for the receipt of enteral vancomycin were assessed, including age, sex, race, hospital, payer status, year of test, timing of test relative to admission, the presence of comorbidities, receipt of antibiotics, and admission in an intensive care unit (ICU) between 24 h before through 48 h after testing.

The demographic characteristics analyzed included age (by quartile), sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, or other), and payer status (public [including Medicaid, Medicare, or other government insurance] versus private [including HMOs, PPOs, and other]). Variables for hospital and time of test were included to allow for analysis of trends over time and to capture hospital-level vancomycin use.

To identify the presence of comorbid chronic conditions, complex chronic condition (CCC) flags, a previously validated diagnostic classification system based on ICD-9-CM codes, were used. CCCs represent diagnoses that are expected to last longer than 12 months and involve an organ system severely enough to require specialty pediatric care and probably hospitalization in a tertiary-care center (15). Categories of CCC include respiratory, renal, gastrointestinal, cardiovascular, hematologic or immunologic, metabolic, genetic or other congenital defects, malignancy, and neuromuscular diagnoses. This method of capturing comorbid diagnoses has been used in prior PHIS studies (2, 16, 17).

To determine whether patients receiving antibiotics at the time of CDI diagnosis were more or less likely to receive enteral vancomycin, antimicrobial therapy around the time of testing (within 1 day before and 2 days after) was assessed. Antibiotic receipt was defined by hospital charge data for one of the following orally or intravenously delivered antibiotic classes or agents: aminoglycoside, penicillin, cephalosporin, macrolide, tetracycline, fluoroquinolone, carbapenem, aztreonam, daptomycin, vancomycin, clindamycin, chloramphenicol, tigecycline, trimethoprim-sulfamethoxazole, spectinomycin, quinupristin-dalfopristin, or colistin. Topical, otic, and ophthalmic antibiotics were excluded, as were any agents that might be used as therapy for CDI (e.g., rifaximin and nitazoxanide). Concurrent receipt of either nitazoxanide or rifaximin was also assessed as an independent predictor of the use of vancomycin. A determination of ICU status was made based on whether room charge data indicated the patient was in an ICU from 1 day before through 2 days after the time of CDI testing.

Trends in CDI outcomes, including mortality, colectomy, length of stay, and total hospital charges were evaluated. All cases were examined for the presence of a procedure code for colectomy using the following ICD-9-CM codes: 45.8 (total intra-abdominal colectomy), 45.81 (laparoscopic total intra-abdominal colectomy), 45.82 (open total intra-abdominal colectomy), 45.83 (other and unspecified total intra-abdominal colectomy), 45.71 (open and other multiple segmental resection of the large intestine), 45.72 (open and other cecectomy), 45.73 (open and other right hemicolectomy), 45.74 (open and other resection of transverse colon), 45.75 (open and other left hemicolectomy), 45.76 (open and other sigmoidectomy), and 45.79 (other and unspecified partial excision of large intestine). Total charges were adjusted for hospital location using the Centers for Medicare and Medicaid Services price/wage index and adjusted to 2011 dollars using the Bureau of Labor Statistics consumer price index for hospital and related services (18).

Statistical analyses.

Summary statistics were constructed using frequencies and proportions for categorical data and means and medians for continuous variables. The temporal trend in vancomycin use was examined using a generalized linear mixed model with a logit link to analyze the relationship between year of testing and receipt of vancomycin. Random intercepts and slopes for each hospital were included to account for within-hospital clustering. Temporal trends in CDI outcomes were also examined using generalized linear models to analyze the relationship between year of testing and the outcome of interest (mortality [logit link] and charges [log link]). In these models, generalized estimating equations (GEEs) were used to adjust for hospital level clustering.

Bivariate and multivariate logistic regression models with GEEs were used to identify potential predictors of vancomycin receipt. All analyses were performed using SAS version 9.3 (SAS Institute, Inc., Cary, NC) with a two-sided P value of <0.05 considered statistically significant.

RESULTS

Antibiotic prescribing.

A total of 6,673 patients were included for analysis (Fig. 1). Fifty-two percent (n = 3,471) of the patients who met the entry criteria were male, and the median age was 6 years (interquartile range, 2 to 13). The most common principal discharge diagnoses included: CDI (n = 1,834; 27.5%); encounter for antineoplastic chemotherapy (n = 318; 4.8%); neutropenia, unspecified (n = 167; 2.5%); dehydration (n = 144; 2.2%); and acute lymphoid leukemia (n = 106; 1.6%). Over the study period, 7% of patients (n = 483) received vancomycin as part of the initial management of their first inpatient episode of CDI. The largest proportion of these patients (n = 230, 47.6%) received vancomycin monotherapy, although a number of patients received more than one agent, either sequentially or concurrently, during the period around CDI testing, including 18.4% (n = 89) who received vancomycin and oral metronidazole, 22.4% (n = 108) who received vancomycin and intravenous metronidazole, and 11.6% (n = 56) who received vancomycin and both oral and intravenous metronidazole.

Fig 1.

Cohort assembly.

The proportion of children who received at least one dose of vancomycin as part of CDI therapy increased significantly over the study period (P = 0.005, Fig. 2) and the rate of change was similar across hospitals. The proportion of children receiving vancomycin during initial CDI management differed significantly between hospitals (range, 0 to 16.7% by hospital; P < 0.001, Fig. 3). Vancomycin utilization differed significantly by U.S. Census Bureau region, with patients in the South (P = 0.002) and West (P = 0.005) less likely to receive vancomycin compared to patients hospitalized in the Northeast. No difference was noted between patients in the Northeast and Midwest (P = 0.61).

Fig 2.

Proportion of children whose initial CDI regimen included enteral vancomycin, grouped by quarter (2006 to 2011).

Fig 3.

Proportion of children whose initial CDI regimen included enteral vancomycin, grouped by hospital.

Predictors of vancomycin use.

Patient demographic and clinical characteristics and the proportion of children who received vancomycin as part of CDI management are displayed in Table 1. In unadjusted analyses, enteral vancomycin was used significantly more often in patients who were older, white, had private insurance, who were tested within 48 h of admission, who were on antibiotics at the time of testing, or who had a gastrointestinal comorbidity.

Table 1.

Demographic and clinical characteristics associated with the use of enteral vancomycin as initial therapy for C. difficile infection

Characteristica	Total no. of subjects	No. (%) of patients that received vancomycin	OR (95% CI)	P
Hospital				<0.001
Region
Northeast	1,071	95 (8.9)	–b
Midwest	1,501	142 (9.5)	1.07 (0.82–1.41)	0.61
South	1,994	116 (5.8)	0.64 (0.48–0.84)	0.002
West	2,107	130 (6.2)	0.68 (0.51–0.89)	0.005
Age (yrs)
1–2	1,946	106 (5.5)	–
3–6	1,471	91 (6.2)	1.15 (0.86–1.53)	0.36
7–13	1,825	146 (8.0)	1.51 (1.17–1.96)	0.002
14–18	1,431	140 (9.8)	1.88 (1.45–2.45)	<0.001
Sex
Male	3,471	248 (7.1)	–
Female	3,202	235 (7.3)	1.03 (0.86–1.24)	0.76
Race
White	3,669	329 (8.7)	–
Black	862	40 (4.6)	0.51 (0.36–0.71)	<0.001
Hispanic	1,205	58 (4.8)	0.53 (0.40–0.70)	<0.001
Other	837	56 (6.7)	0.75 (0.56–1.01)	0.05
Insurance
Private	3,597	303 (8.4)	–
Public	3,076	180 (5.8)	0.68 (0.56–0.82)	<0.001
Yr of test
2006	1,124	77 (6.9)	–
2007	1,227	68 (5.5)	0.80 (0.57–1.12)	0.19
2008	1,317	94 (7.1)	1.05 (0.77–1.43)	0.78
2009	1,296	83 (6.4)	0.93 (0.68–1.28)	0.66
2010	1,207	108 (9.0)	1.34 (0.99–1.81)	0.06
2011	502	53 (10.6)	1.61 (1.11–2.32)	0.01
Timing of test
>48 h after admission	3,108	179 (5.8)	–
≤48 h after admission	3,565	304 (8.5)	1.53 (1.26–1.85)	<0.001
Antibiotics around test*
No	1,991	178 (8.9)	–
Yes	4,682	305 (8.5)	0.71 (0.59–0.86)	<0.001
Nitazoxanide
No	6,602	474 (7.18)	–
Yes	71	9 (12.68)	1.88 (0.93–3.80)	0.08
Rifaximin
No	6,651	480 (7.22)	–
Yes	22	3 (13.64)	2.03 (0.60–6.88)	0.26
ICU around test*
No	5,625	396 (7.0)	–
Yes	1,048	87 (8.3)	1.20 (0.94–1.52)	0.15
Any CCC
No	2,120	155 (7.3)	–
Yes	4,553	328 (7.2)	0.98 (0.81–1.20)	0.87
No. of CCC
0	2,120	155 (7.3)	–
1	3,113	227 (7.3)	1.00 (0.81–1.23)	0.98
≥2	1,440	101 (7.0)	0.96 (0.74–1.24)	0.74
Cardiovascular CCC	641	40 (6.2)	0.84 (0.60–1.17)	0.31
Gastrointestinal CCC	700	105 (15.0)	2.61 (2.07–3.30)	<0.001
Hematologic/immunologic CCC	422	35 (8.3)	1.17 (0.82–1.68)	0.39
Malignancy CCC	1,973	88 (4.5)	0.51 (0.40–0.65)	<0.001
Metabolic CCC	598	44 (7.4)	1.02 (0.74–1.41)	0.91
Neuromuscular CCC	997	69 (6.9)	0.95 (0.73–1.23)	0.68
Other congenital/genetic defect CCC	545	46 (8.4)	1.20 (0.87–1.65)	0.26
Renal CCC	136	12 (8.8)	1.25(0.68–2.27)	0.47
Respiratory CCC	439	26 (5.9)	0.80 (0.53–1.20)	0.27

^aICU, intensive care unit; CCC, complex chronic condition. *, “Around test” refers to the period 24 h before through 48 h after C. difficile testing.

^b–, a dash indicates that this value was used as the reference value.

In multivariate models adjusting for hospital-level clustering and potential confounders, several covariates remained independent predictors of vancomycin receipt (Table 2). Patients 7 to 13 years and 14 to 18 years of age were more likely to receive enteral vancomycin. Children who were black or Hispanic were less likely to receive vancomycin, even after controlling for geographic region. Children with gastrointestinal conditions were more likely, while children with malignancies were less likely to receive vancomycin. Patients in an ICU at the time of testing were more likely to receive vancomycin, as were patients whose tests were performed within 48 h of admission. Finally, for each successive year of the study the odds of vancomycin receipt increased by 10%.

Table 2.

Independent predictors of the use of enteral vancomycin as initial therapy for C. difficile infection

Predictora	Adjusted OR (95% CI)	P
Age (yrs)
1–2	–b
3–6	1.13 (0.92–1.40)	0.25
7–13	1.40 (1.11–1.76)	0.005
14–18	1.57 (1.13–2.18)	0.008
Race
White	–
Black	0.53 (0.39–0.73)	<0.001
Hispanic	0.63 (0.47–0.84)	0.002
Other	0.79 (0.51–1.24)	0.31
Time of test (yrs)	1.10 (1.03–1.18)	0.006
Test ≤48 h after admission	1.28 (1.02–1.62)	0.006
ICU around testc	1.37 (1.05–1.80)	0.02
Gastrointestinal CCC	2.01 (1.44–2.81)	<0.001
Malignancy CCC	0.57 (0.36–0.89)	0.01

^aICU, intensive care unit; CCC, complex chronic condition.

^b–, a dash indicates that this value was used as the reference value.

^c“Around test” refers to the period 24 h before through 48 h after C. difficile testing.

Outcomes for children with CDI.

Among all patients, the odds of death decreased by 12% per year over the study period, with mortality rates dropping from 3.9% in 2006 to 2.6% in 2011 (P = 0.02). There was no change in the number of patients requiring ICU-level care (P = 0.88). No significant change for any other outcome, including the rate of colectomy (P = 0.92), length of stay (P = 0.50), or total hospital charges (P = 0.86) was noted during the study period.

DISCUSSION

The management of pediatric CDI is based largely on an extrapolation of findings from adult studies, since clinical trials including children are lacking. Whether pediatric providers have adopted adult practice patterns, including recommended indications for the use of enteral vancomycin, is unknown. Prior pediatric studies that have included information regarding CDI-directed pharmacotherapy have yielded widely disparate estimates of vancomycin use, ranging from 3.5 to 46%, and more recent data are lacking (2, 19, 20). Our study demonstrates that the use of enteral vancomycin as first-line management of pediatric CDI is increasing, even while rates of severe outcomes remain low. Furthermore, certain characteristics including older age, white race, intensive care status, early CDI testing, and gastrointestinal comorbidity have emerged as important predictors of the use of vancomycin.

Although there have been no direct comparisons of vancomycin and metronidazole for CDI treatment in children, our study suggests that providers are increasingly turning to vancomycin. The limitations of administrative data do not allow definitive conclusions about the reason for this trend, although there are several potential explanations. Increasing vancomycin use could reflect a concurrent increase in CDI severity, although we found no significant change in the number of patients requiring ICU-level care and no increase in serious CDI outcomes. A second possibility is an increase in the number of patients at risk for severe CDI. This explanation also seems unlikely, since adjustment for comorbidities that might be associated with an increased risk of severe disease did not attenuate the observed increase in vancomycin use. PCR testing for C. difficile became more widely available toward the end of our study period. However, a more sensitive diagnostic test would not be expected to affect the choice of therapeutic agent. A final possibility is that the trend toward more vancomycin use reflects evolving prescriber preferences. This hypothesis may explain the observed association between specific underlying diagnoses and vancomycin utilization. In addition to an overall increase in use, we identified significant variability in vancomycin utilization between hospitals, even after adjusting for patient characteristics. Although variation in CDI management has not been studied previously, unexplained hospital-level variability in antimicrobial use has been well described for a variety of pediatric conditions (21–24).

Demographic characteristics associated with increased vancomycin use included older age and white race. Reasons for greater vancomycin use in older children might include greater comfort with vancomycin in older populations or more severe disease in older children. One recent pediatric study found an association between increasing age and increased severity of CDI, although this finding was no longer statistically significant after children <1 year of age were removed from the sample (19). The impact of race on CDI epidemiology and variations in care has not been explored; our findings suggest that children who are black or Hispanic are less likely to receive vancomycin. These differences remained statistically significant, even after accounting for differences in geographic location. The reasons for this apparent racial and ethnic variation in vancomycin use are unclear but may reflect practice variation based on patient socioeconomic status. Clinicians' thresholds for treating patients with vancomycin, a more costly agent, may differ based on a patient's ability to afford this drug. Payer status was noted to be associated with vancomycin use on univariate analysis; however, due to strong colinearity with race, did not enter into our final model or fully explain the variation in vancomycin use on the basis of race. One recent study demonstrated an increase in risk of CDI among white children, although black children had an increased risk of death, longer hospital stay, and higher hospitalization charges (3). Despite the suggestion that black children with CDI have worse outcomes, in our study they were significantly less likely to receive vancomycin. Future studies aimed at exploring the relationship between outcomes and therapy in black and Hispanic children are warranted.

We found increased vancomycin use among patients with a gastrointestinal comorbidity and decreased use among patients with malignancies. A number of diagnoses are captured by the gastrointestinal CCC flag used in the present study, including congenital anomalies (e.g., tracheoesophageal fistula and Hirschsprung's disease), liver diseases, and inflammatory bowel disease (IBD). Rates of CDI are higher among patients with IBD and mortality rates, hospital length of stay, and rates of lower gastrointestinal surgery and endoscopy are also increased compared to patients without IBD (3, 25, 26). Pediatric patients with IBD have a higher rate of CDI recurrence and tend to have a more severe IBD clinical course (27). Despite these findings, whether patients with underlying gastrointestinal comorbidities should be managed with vancomycin as first-line therapy for mild-to-moderate CDI remains uncertain. Recent data suggest that risk factors for severe CDI may be different in patients with IBD, suggesting that preferred approaches to management may differ for these patients (28). To date, however, there are no prospective data comparing metronidazole and vancomycin among patients with IBD or any other gastrointestinal diseases. Children with underlying malignancy are also at increased risk for CDI, and at least one study suggests that these patients may also have more severe disease (19, 29). Interestingly, we found that despite this potential for severe illness, these patients are less likely to receive vancomycin. This finding may reflect greater concern for neutropenic colitis among these patients and the preference for an empirical antibiotic with anaerobic coverage. Alternatively, patients with malignancies may not necessarily be at risk for more severe disease or worse outcomes, as suggested by a more recent study (30). Qualitative studies aimed at understanding the rationale for prescribing patterns among pediatric subspecialists would be helpful to elucidate observed antibiotic choices for selected patient populations.

The PHIS data set has several limitations. First, it does not include laboratory results that would allow for the identification of patients with severe disease using IDSA/SHEA criteria. In the absence of these data, we chose to include ICU status around the time of CDI testing as a marker for severe disease, which was associated with an increased likelihood of vancomycin use. Second, PHIS does not contain outpatient information, which may have influenced the finding that patients who underwent CDI testing within 48 h of admission were more likely to receive vancomycin. We hypothesize that these patients represent children with CDI who required an escalation of therapy due to refractory outpatient disease. Additional limitations of the present study include generalizability to pediatric patients not managed in freestanding children's hospitals and the lack of laboratory results to confirm CDI diagnoses. To avoid potential misclassification bias, we used a previously validated case definition and limited our analysis to the first admission for which a patient had a CDI discharge diagnosis (2, 14). Even using this strategy, we may have included some patients who had prior episodes of CDI that did not require inpatient admission.

Existing recommendations, including a recent policy statement from the AAP Committee on Infectious Diseases, suggest that on the basis of efficacy, cost, and the goal of antimicrobial stewardship, oral metronidazole should remain the drug of choice for the initial treatment of children with CDI (8). In the absence of pediatric data demonstrating superiority of vancomycin over metronidazole in the management of CDI and with evidence that rates of severe outcomes have remained stable, the observed national increase in vancomycin use may be unjustified. There are specific patient characteristics, including race and ethnicity, that are associated with vancomycin receipt, and these populations should be the initial focus of efforts aimed at understanding variation in CDI management. As rates of CDI continue to rise, a comparative examination of agents used to treat CDI in children and a better understanding of current practice variations and prescriber preferences is urgently needed.