Clostridioides difficile Infections in Inpatient Pediatric Oncology Patients: A Cohort Study Evaluating Risk Factors and Associated Outcomes

Abstract

Background

Clostridioides difficile infection (CDI) is a significant source of morbidity in pediatric cancer patients. Few reports to date have evaluated risk factors and short-term outcomes for this population.

Methods

We retrospectively evaluated pediatric oncology admissions at St Louis Children’s Hospital from 2009 to 2018. All inpatient cases of diagnosed initial CDI were identified. We aimed to investigate the prevalence of CDI and associated risk factors, including coadmission with another patient with CDI, and to evaluate short-term outcomes including length of stay and delays in subsequent scheduled chemotherapy.

Results

Review of 6567 admissions from 952 patients revealed 109 CDI cases (11.4% of patients). Patients with leukemia or lymphoma, compared to those with solid tumors, were more likely to have CDI (odds ratio [OR], 3 [95% CI, 1.4–6.6], and 3 [95% CI, 1.3–6.8], respectively). Autologous hematopoietic stem cell transplant (HSCT) was also a risk factor (OR, 3.5 [95% CI, 1.7–7.4]). Prior antibiotic exposure independently increased the risk for CDI (OR, 3.0 [95% CI, 1.8–4.8]). Concurrent admission with another patient with CDI also significantly increased the risk (OR, 84.7 [95% CI, 10.5–681.8]). In contrast to previous reports, exposure to acid-suppressing medications decreased the risk for CDI (OR, 0.5 [95% CI, .3–.7]). CDI was associated with increased length of stay (mean difference, 8 days [95% CI, 4.6–11.4]) and prolonged delays for subsequent chemotherapy (mean difference, 1.4 days [95% CI, .1–2.7]).

Conclusions

CDI in pediatric oncology patients significantly prolongs hospitalization and delays chemotherapy treatment plans. Interventions to control CDI will improve the care of pediatric oncology patients.

Clostridioides difficile infection (CDI) is associated with prolonged delays in subsequent chemotherapy administration in pediatric oncology patients. Exposure to other patients with CDI significantly contributes to infection risk. These findings define opportunities to improve the care of these patients.

Clostridioides difficile is a spore-forming bacillus that colonizes the large intestine. If it is able to proliferate and produce toxins, it commonly results in an inflammatory diarrheal illness. Consequences can include pseudomembranous colitis, toxic megacolon, hypovolemia, septic shock, and death, especially in at-risk hosts [1]. Oncology patients have been shown to be particularly at risk. Rates vary by malignancy, with hematopoietic stem cell transplant (HSCT) recipients having the highest rates of infection, a reflection of the intensity of antibiotic exposures, chemotherapy, and immunocompromised state [2–5]. In adults with malignancies receiving chemotherapy, the frequency of CDI has been shown to be as high as 16% and as high as 20% in HSCT recipients.

Pediatric oncology patients have also been shown to be at increased risk for developing CDI; however, the data are more limited. Some studies have shown incidence rates of CDI in pediatric oncology patients to be 15 times higher compared to the general pediatric population [6–8]. Special groups appear to be at even higher risk, with up to 17% of HSCT recipients developing CDI [9]. Other risk factors for CDI in pediatric oncology patients include prolonged length of stay (LOS), use of multiple antibiotics, receipt of chemotherapy, and immunosuppression [10–12]. However, reports to date have been limited by sample size and have not included patient courses throughout all stages of therapy to fully characterize their risk for infection.

The severity of effect of CDI on the care of pediatric oncology patients has not been extensively studied. Studies in pediatric patients with acute myeloid leukemia demonstrate that few patients develop complicated disease; however, there are few data characterizing the impact of CDI on the outcomes in all pediatric oncology patients and its impact on the treatment of their malignancies [13].

METHODS

The purpose of this study was to evaluate the prevalence and risk factors for CDI in inpatient pediatric oncology patients and to assess their associated outcomes compared to pediatric oncology patients without CDI at a large academic pediatric hospital.

Patient Population

In this retrospective cohort study, all patients with a diagnosed malignancy admitted to the St Louis Children’s Hospital (SLCH) oncology service between July 2009 and February 2018 were included. All oncology patients are admitted to the same unit of the hospital; thus, admissions to the unit comprised the screened population and initial study cohort. All admissions were included with common admission indications for oncology patients including scheduled chemotherapy infusions, fevers and infections, and need for pain control or hydration. Cancer patients were screened positive with any International Classification of Diseases, Ninth Revision or Tenth Revision codes for a malignant neoplasm using billing codes. A manual chart review was performed to confirm the cancer diagnosis to establish the final cohort. Only patients with confirmed malignancies were included. Cases were defined as any admission with CDI. Controls were defined as any admission without CDI. For patients with multiple admissions, we collected data from all admissions during the study period. For patients who experienced CDI, all admissions after the first episode of CDI were excluded, as previous CDI is a risk factor for subsequent CDI.

For HSCT patients, the date and type of initial transplant procedure was documented, and confirmed using the SLCH bone marrow transplant program’s institutional log. In the event a patient was transplanted multiple times, the first transplant date and type were used for analysis. A transplant admission was defined as any admission in which an HSCT occurred or any subsequent admissions for those patients.

Risk Factors

To evaluate medications as CDI risk factors, we created dichotomous variables for antibiotic exposure (all intravenous [IV] cephalosporins, carbapenems, or fluoroquinolones) and acid suppression exposure (all proton pump inhibitors [PPIs] and H2 receptor antagonists). Vancomycin and metronidazole were excluded due to indications for CDI treatment. All other high-risk IV antibiotics represented a very small percentage of antibiotic use. Exposure was defined as any administration of the medications between admission and time of CDI for admissions with CDI, and any administration at any time point for admissions without CDI.

As CDI can be acquired by contact with infectious patients, we sought to evaluate the relationship between existing infections on the unit and new infections. For each admission, we created a dichotomous variable, CDI exposure, to evaluate the risk of CDI acquisition attributable to coadmission on the oncology unit during the infectious period of another oncology or transplant admission with CDI. Infections in nononcology patients were not included. Patients with a positive C difficile assay were considered infectious for up to 14 days following initial diagnosis or until the time of discharge [14]. Any admissions that overlapped with the infectious period of another oncology patient with CDI were considered positive for CDI exposure.

Outcome Measures

CDI was defined as a positive laboratory result for detection of C difficile toxin in a patient with liquid stool. Formed stools are rejected by the clinical laboratory, so tested specimens were restricted to patients with diarrhea. Laboratory testing was performed by toxin detection enzyme-linked immunosorbent assay (ELISA) prior to 2011, by glutamate dehydrogenase (GDH) screen with confirmatory C difficile toxin B polymerase chain reaction (PCR) assay (Cepheid Xpert) from 2011 to May 2017, and by toxin A/B ELISA thereafter (TechLab C difficile Tox A/BII).

Chemotherapy administration dates were obtained from patient chemotherapy administration roadmaps, which are date-populated flowsheets documenting all administered treatment for the SLCH oncology program.

Chemotherapy delay was assessed both as a dichotomous and continuous variable. We first identified which patients and admissions had chemotherapy scheduled following discharge. For each of these admissions, we assessed whether chemotherapy was administered on the day it was due or was delayed, and if so, the number of days by which therapy was delayed. The projected date of planned chemotherapy was determined by the patient chemotherapy administration treatment document (“roadmap”). For a small proportion of admissions, therapy was truncated prematurely following the admission. These cases were included in dichotomous delay analysis but were excluded from continuous analysis.

Statistical Analysis

SAS version 9.4 software (SAS Institute, Cary, North Carolina) was used to perform all statistical analysis. Both univariate and multivariate generalized linear models with repeated measures were applied to examine the association between outcomes and risk factors. Logit and identity link functions were used for binary and continuous outcomes, respectively. As CDI increases the likelihood of subsequent CDI, all patient admissions were censored after the first episode of CDI to eliminate this influence. A generalized estimating equation approach was used to estimate the model parameters assuming an exchangeable covariance structure, and to account for multiple admissions from each patient. Stepwise selection was used in the multivariate analyses, where a significance level of 0.3 was required to allow a risk factor into the model, and a significance level of 0.15 was required for a risk factor to stay in the model. The final model included risk factors with a significance level of 0.05. Multiple comparison across multiple levels of a specific risk factor was adjusted by Tukey-Kramer method.

A control chart was created to evaluate variation in rate of CDI on the pediatric oncology unit. The average rate ($\overline{x}$) was calculated using total number of CDI cases per number of admissions to the unit and recalculated each time the testing modality changed. The standard deviation (SD) was calculated using number of admissions per month to create varying SD, corrected for admission volume. The upper control limit was set at $\overline{x}+2SD$.

RESULTS

During the study period, we identified 952 pediatric oncology patients with 6567 admissions. Tables 1 and 2 show the patient characteristics and admission characteristics by CDI status. Age at the time of index admission ranged from 0.5 months to 24 years with a median age of 7.7 years. The number of admissions per patient was highly skewed, ranging from 1 to 40 admissions with a median of 5 admissions per patient. Mean and median LOS was 6.4 and 3.2 days, respectively.

Table 1.

Characteristics of Pediatric Oncology Patients During the Study Period

Characteristic	Patients (N = 952)
Age group, ya
0–1	128 (13.4%)
2–4	214 (22.5%)
5–9	218 (22.9%)
10–14	189 (19.9%)
15–19	181 (19%)
≥20	22 (2.3%)
Sex
Male	533 (56%)
Female	419 (44%)
Race/ethnicity
African American	132 (13.9%)
Asian	9 (1%)
White	762 (80%)
Other/unknown	49 (5.1%)
Diagnosis
Brain tumor	177 (18.6%)
Histiocytosis	16 (1.7%)
Leukemia	284 (29.8%)
Lymphoma	132 (13.9%)
Solid tumor	337 (35.4%)
Other	6 (0.6%)
HSCT	159 (16.7%)
Allogeneic	78 (49.1%)
Autologous	81 (50.9%)

Data are presented as No. (%).

Abbreviation: HSCT, hematopoietic stem cell transplant.

^aPatient-level age calculated as years between first admission date and date of birth.

Table 2.

Characteristics at Admissions During the Study Period by Clostridioides difficile Infection Status

Characteristic	Total	With CDI	Without CDI
No. of admissions	6567 (100%)	109 (1.7%)	6458 (98.3%)
Sex
Male	3669 (55.9%)	56 (51.4%)	3613 (55.9%)
Female	2898 (44.1%)	53 (48.6%)	2845 (44.1%)
Race
African American	883 (13.5%)	14 (13%)	869 (13.5%)
Asian	56 (0.9%)	2 (1.8%)	54 (0.8%)
White	5338 (81.5%)	88 (81.5%)	5250 (81.5%)
Other	271 (4.1%)	4 (3.7%)	267 (4.2%)
Diagnosis group
Brain tumor	996 (15.2%)	10 (9.2%)	986 (15.3%)
Histiocytosis	48 (0.7%)	1 (0.9%)	47 (0.7%)
Leukemia	1853 (28.2%)	54 (49.5%)	1799 (27.9%)
Lymphoma	649 (9.9%)	16 (14.7%)	633 (9.8%)
Solid tumor	2981 (45.4%)	27 (24.8%)	2954 (45.7%)
Other	40 (0.6%)	1 (0.9%)	39 (0.6%)
Transplant status
Allogeneic	406 (6.2%)	15 (13.8%)	391 (6.1%)
Autologous	337 (5.1%)	14 (12.8%)	323 (5%)
No transplant	5824 (88.7%)	80 (73.4%)	5744 (88.9%)
CDI exposurea
No	2852 (43.5%)	1 (0.9%)	2851 (44.2%)
Yes	3703 (56.5%)	108 (99.1%)	3595 (55.8%)
Antibiotic prior to CDIb
No	4048 (61.6%)	31 (28.4%)	4017 (62.2%)
Yes	2519 (38.4%)	78 (71.6%)	2441 (37.8%)
Acid suppression prior to CDI
No	4422 (67.3%)	76 (69.7%)	4346 (67.3%)
Yes	2145 (32.7%)	33 (30.3%)	2112 (32.7%)
Chemotherapy delayc
No	2724 (61.5%)	32 (49.2%)	2692 (61.7%)
Yes	1703 (38.5%)	33 (50.8%)	1670 (38.3%)
Aged
No.	6567	109	6458
Mean, y (SD)	9.3 (5.9)	9.5 (5.9)	9.2 (5.9)
Length of stay prior to CDIe
No.	6552	109	6443
Mean, d (SD)	6.2 (10.6)	8.1 (17.4)	6.2 (10.5)
Length of stayf
No.	6552	109	6443
Mean, d (SD)	6.4 (11)	18 (24.7)	6.2 (10.5)
Chemotherapy delayg
No.	4399	63	4336
Mean, d (SD)	2.5 (5.2)	4.2 (5.8)	2.5 (5.2)

Abbreviations: CDI, Clostridioides difficile infection; SD, standard deviation.

^aDefined as coadmission during the infectious period of an admission with CDI.

^bCephalosporin, carbapenem, or fluoroquinolone.

^cAmong 6567 admissions, 2140 missing records on chemotherapy delay.

^dAdmission-level age calculated as years between each admission date and date of birth.

^eLength of stay prior to CDI calculated as days between CDI diagnosis date and admission date.

^fAmong 6567 admissions, 15 missing records on length of stay.

^gAmong 6567 admissions, 2168 missing records on days in chemotherapy delay.

Risk Factors for CDI

We identified 109 cases of CDI from 109 unique individuals (Table 2), representing 11.4% of patients and 1.7% of admissions. The vast majority of infections were classified as healthcare associated, by Centers for Disease Control and Prevention (CDC) definitions [15]. The mean time from admission to CDI was 8.1 days. Median time to discharge following CDI diagnosis was 5.9 days. Nearly half of CDIs occurred in admissions of patients with leukemia (54/109), and 25% occurred in admissions of patients with solid tumors (27/109). Twenty-nine percent of infections occurred in patients who had received an HSCT (29/109; 14 autologous and 15 allogeneic).

Age, sex, and race were not associated with CDI. On multivariate analysis, leukemia and lymphoma diagnoses, compared to solid tumor, were significantly associated with CDI (odds ratio [OR], 3.0 [95% confidence interval {CI}, 1.4–6.6], and OR, 3.0 [95% CI, 1.3–6.8], respectively). Compared to no transplant, autologous transplant was associated with CDI (OR, 3.5 [95% CI, 1.7–7.4]), whereas allogeneic transplant was not (OR, 1.4 [95% CI, .6–3.3]). Antibiotic exposure was also associated with CDI admission (OR, 3.0 [95% CI, 1.8–4.8]). Acid suppression medications were negatively related to CDI (OR, 0.5 [95% CI, .3–.7]) (Table 3). CDI exposure was strongly associated with CDI (OR, 84.7 [95% CI, 10.5–681.8]). During the study period, the proportion of CDI admissions clustered, resulting in a monthly incidence greater than the upper control limit on 6 occasions, occurring during winter and spring months (Figure 1).

Table 3.

Univariate and Multivariate Logistic Regression Model With Repeated Measures Assessing Risk Factors for Clostridioides difficile Infection

Risk Factors	Univariate		Multivariatea
	Odds Ratio (95% CI)	P Valueb	Odds Ratio (95% CI)	P Valueb
Age	1 (.98–1.04)	.669
Sex, female vs male	1.2 (.8–1.7)	.322	…
Race
White	---		…
Asian	2.3 (.5–9.7)	.449	…
Black	0.9 (.5–1.9)	.994	…
Other	0.9 (.2–3.2)	.991	…
Length of stay prior to CDIc	1 (1.001–1.02)	.040	…
Diagnosis group
Solid tumor	…		…
Leukemia	3.3 (1.7–6.2)	<.001	3 (1.4–6.6)	<.001
Lymphoma	2.9 (1.3–6.5)	.004	3 (1.3–6.8)	.002
Histiocytosis	2.3 (.1–42.6)	.961	1.9 (.1–47.3)	.994
Other	3.2 (.8–12.9)	.185	3.1 (.3–28.6)	.71
Brain tumor	1.2 (.4–3.1)	.999	1.5 (.6–4.1)	.858
Transplant during admission
No transplant	…		…
Autologous transplant	3.1 (1.6–6.1)	<.001	3.5 (1.7–7.4)	<.001
Allogeneic transplant	2.7 (1.4–5.2)	.002	1.4 (.6–3.3)	.626
CDI exposured
Yes vs no	96.2 (11.2–823.6)	<.001	84.7 (10.5–681.8)	<.001
Antibiotic prior to CDIe
Yes vs no	4.4 (2.9–6.7)	<.001	3 (1.8–4.8)	<.001
Acid suppression prior to CDI
Yes vs no	1 (.6–1.4)	.834	0.5 (.3–.7)	.002

Abbreviations: CDI, Clostridioides difficile infection; CI, confidence interval.

^aMultivariate repeated measure logistic regression stepwise selection results.

^bMultiple comparison was adjusted by Tukey-Kramer method.

^cLength of stay prior to CDI calculated as days between CDI diagnosis date and admission date.

^dDefined as coadmission during the infectious period of an admission with CDI.

^eCephalosporin, carbapenem, or fluoroquinolone.

Figure 1.

Control chart illustrating monthly variation in rate of Clostridioides difficile infection on the oncology and transplant unit, with 2 standard deviations (2σ) upper control limit, adjusted for number of admissions per month. Abbreviations: CDI, Clostridioides difficile infection; ELISA, enzyme-linked immunosorbent assay; GDH, glutamate dehydrogenase; PCR, polymerase chain reaction.

Outcome Variables Linked to CDI

CDI was strongly associated with increased LOS compared to admissions without CDI, with a mean difference of 8 days (95% CI, 4.6–11.4 days) on multivariate analysis. CDI was also significantly linked to the length of chemotherapy delay. Although the prevalence of chemotherapy delay did not differ between admissions with CDI vs those without (OR, 1.6 [95% CI, 1.0–2.5]), in admissions with CDI, the average length of chemotherapy delay was 1.4 days (95% CI, .1–2.7 days) longer than in admissions without CDI (Table 4).

Table 4.

Multivariate Generalized Linear Model With Repeated Measures Assessing Outcomes in Clostridioides difficile Infection (CDI) Admissions Versus Admissions Without CDI

Outcome	Admissions With CDI vs Without CDI	P Value
Length of stay, mean difference (95% CI)a	8 (4.6–11.4)	<.001
Chemotherapy delay, OR (95% CI)b	1.6 (1.0–2.5)	.052
Days of chemotherapy delay, mean difference (95% CI)c	1.4 (.1–2.7)	.035

Abbreviations: CI, confidence interval; CDI, Clostridioides difficile infection; OR, odds ratio.

^aAmong 6567 admissions, 15 missing records on length of stay.

^bAmong 6567 admissions, 2140 missing records on chemotherapy delay.

^cAmong 6567 admissions, 2168 missing records on days in chemotherapy delay.

DISCUSSION

Our study evaluating >950 total patients including 109 with CDI and spanning 9 years, represents one of the largest cohort studies of CDI in the pediatric oncology population. While mortality is rare, it is important to understand how the morbidity in children differs from that of adults. Our study demonstrates that admissions in which CDI occurs are significantly prolonged compared to those without CDI. Therefore, CDI significantly impacts the care of pediatric cancer patients.

To our knowledge, this study is the first to evaluate chemotherapy-related outcomes following CDI in all pediatric oncology and HSCT patients. Although admissions with CDI did not result in more chemotherapy delays, we found that chemotherapy delays following CDI admissions were significantly longer than following non-CDI admissions. This may have major implications for the long-term therapy for these patients. Pediatric oncology therapy is highly regimented, often with compressed schedules of chemotherapy, and variance from the planned therapy has been shown to impact disease-free survival [16, 17]. Prospective evaluations may allow us to further define the factors that contribute to these delays, and could point to interventions that would minimize the impact on the patient’s cancer therapy. One potential explanation for why frequency of delays did not differ is that short delays due to scheduling or family convenience were relatively common in both groups.

Similar to previous studies, we found antibiotics and underlying malignant disease to be major factors associated with CDI [6, 8, 11, 18–20]. Antibiotic usage in our program follows standard processes and varied little over the course of this study. This consistency allowed us more confidence in the impact of antibiotic use on the incidence of CDI. Our service also has a consistent presence of midlevel providers, which increases consistency in patient care and prescribing practices. Diagnoses of leukemia or lymphoma have both been described previously as risk factors for CDI. These patients are more severely immunosuppressed, which likely contributes to this risk. Surprisingly, although HSCT was a risk factor for CDI, when stratified by cell source, allogeneic transplant was not found to be an independent risk factor for CDI. Of note, allogeneic HSCT recipients are the only patient population who receive ciprofloxacin prophylaxis, which is initiated upon discharge from the transplant procedure. Recent studies have suggested that levofloxacin prophylaxis reduces the risk of CDI [21]. Thus, our use of ciprofloxacin may have influenced the prevalence of CDI in the allogeneic HSCT patient population. Due to the intensive preparatory regimens administered during autologous transplant procedures, patients experience more mucositis in these procedures compared with allogeneic transplant procedures, where there is an increasing emphasis of reduced-intensity conditioning [22]. This distinction between the 2 transplant modalities may contribute to the differences in CDI risk, as mucosal barrier injury from intensive preparative regimens can increase the risk for infection, fever, and antibiotic use.

CDI exposure was found to be strongly predictive of CDI. While this has been investigated in adults, this has not been reported in pediatric oncology patients. A CDI admission was associated with CDI exposure, with a remarkable OR of 84.7 (95% CI, 10.5–681.8), which is an order of magnitude greater than any other risk factor assessed. We believe the magnitude of influence of CDI exposure has broad implications for future prevention techniques. At our institution, all patients with diarrhea are placed on contact isolation precautions. Patients with CDI remained on isolation until discharge, and rooms were terminally cleaned by environmental services with bleach products. Despite these efforts, improved strategies to limit the spread of CDI are needed if a decrement of CDI can be expected from these patients. Clusters of CDI cases also occurred primarily in winter and early spring, consistent with prior reports of seasonality [23]. These seasons commonly have higher patient volumes, and further studies focused on nurse assignment and patient load may help evaluate if caregiver transmission is a contributing factor.

Acid suppression was not shown to be a risk factor for CDI and instead was shown to be protective. The association between acid suppression and CDI is controversial. Our findings are in contrast to many previous studies including a recent systematic review implicating acid suppression as a risk factor for CDI [24]. One possible explanation for this finding is that H2 antagonists and PPIs were analyzed as a single cohort for statistical analysis. Unfortunately, the use of H2 antagonists and PPIs was too rare to perform stratified analysis with our data set. The recent systematic review found PPIs to be predictive, whereas H2 antagonists were not. Other variation may be due to differences in patient population, as perhaps pediatric patients are less susceptible to this risk.

Our study has several limitations related to its retrospective nature. While our study evaluated 8.5 years of admissions including >6000 individual admissions, the number of CDI cases analyzed is still relatively small. This made evaluation of some risk factors, such as individual antibiotics or concurrent infections, impossible. Due to these limitations, antibiotic exposure was evaluated as a dichotomous variable, which does not account for duration of exposure. Available data were limited to inpatient medications and risk factors only. We acknowledge that there may be outpatient exposures that influence the risk of CDI. We unfortunately did not have the ability to assess this as ambulatory surveillance of CDI was not done routinely; thus, any description of positive CDI testing in the ambulatory setting would be an underrepresentation. Future studies are needed to gain insight into the impact of outpatient CDI on the overall incidence of CDI. Also due to the retrospective nature of this study, details of patient symptoms, CDI treatment course, and reason for chemotherapy delay were largely unavailable as they were not present in the medical record. However as formed stools are rejected by the clinical laboratory for C difficile testing and testing is only sent if there is clinical suspicion, we could safely assume all patients had diarrhea, although we could not rule out other alternative or concomitant causes of some of the episodes. Similarly, we acknowledge that infants and young children have a higher prevalence of colonization and less commonly have symptomatic CDI [25]. Pediatric oncology patients as a group also have colonization rates several-fold higher than healthy children [26]. Nonetheless, they were included in the analysis, since a positive test represented evidence of suspicion and in the pediatric oncology patient, a positive test would be treated as CDI and not presumed to be colonization.

Only 5 episodes of CDI were community acquired by CDC definitions. Pediatric oncology patients have frequent hospitalizations and frequent clinic visits. As our outpatient oncology clinic is located on the same floor as the inpatient unit, we felt this exposure to the hospital made “community acquired” an inaccurate characterization for these patients. Thus, these patients were included in our analysis.

Additionally, CDI diagnostic modalities changed during the course of this study; 2-step detection with GDH screening followed by PCR detection confirmation was used during the majority of the study period, but toxin detection ELISA was used before 2011 and after May 2017. These changes in modality for CDI detection could affect the sensitivity of detection of C difficile. PCR alone has greater sensitivity to detect C difficile compared to GDH screening with PCR confirmation; however, this also may increase detection of asymptomatic colonization as opposed to true CDI [27]. With the addition of GDH screening, fewer asymptomatic carriers are identified. As such, a clear difference between CDI incidence with toxin ELISA and GDH/PCR techniques has not been identified and we believe it did not have a major impact on the study findings; this small change in detection is illustrated in Figure 1. Finally, our study included all admissions to the oncology service including short admissions for observation and prolonged admissions for significant illness. Further characterization for reasons for admission could not be delineated from the chart. In the future, stratification by indication for hospitalization could reveal new unappreciated risk factors potentially providing new opportunities for intervention.

In conclusion, our study showed that CDI in pediatric cancer patients significantly impacts hospital LOS and impedes the adherence to scheduled chemotherapy plans. Larger, multi-institutional prospective trials are needed to further evaluate risk factors and to test interventions that may reduce infection rates. Additionally, coadmission with patients with CDI is highly associated with new cases of CDI. Improvements in and adherence to strict infection control practices will hopefully contribute to the reduction of infection rates and enhancement of patient care for this population.

Notes

Acknowledgments. We gratefully acknowledge the contributions of Gregory Storch, Phillip Tarr, and Jennifer Tappenden. We thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St Louis, Missouri, for the use of the Biostatistics Shared Resource, which provided statistical support.

Financial support. This work was supported by the Children’s Discovery Institute of Washington University and St Louis Children’s Hospital (grant number MC-MI-F-2019–801); and by the National Cancer Institute Cancer Center (Support Grant number P30 CA091842).

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.