Clostridium difficile is a gram-positive, spore-forming, anaerobic bacillus that can colonize the gastrointestinal tract and can lead to C difficile infection (CDI). CDI has a wide variation in severity, ranging from asymptomatic colonization to severe diarrhea, pseudomembranous colitis, toxic megacolon, bowel perforation, and death. In recent years, the incidence of CDI, number of hospitalizations, associated deaths, and severity in adults have been increasing.1,2
The increased rate and severity of CDI have been observed in conjunction with the description of hypervirulent strains of C difficile that are resistant to antibiotics (such as fluoroquinolones and third-generation cephalosporins), that hypersecrete toxins A and B, and that secrete an additional toxin known as binary toxin.3,4 The hypervirulent strain known as North American Pulsed Field type 1 (NAP1) occurs in 19.4% of pediatric patients and is associated with more complications in children.5- 8 However, much of the increased frequency and severity of CDI may be due to factors other than the emergence of hypervirulent strains.
In previous nationwide studies, CDI rates have been the highest in elderly and urban populations.1,9 More recent reports describe severe CDI in children and young, previously healthy patients without hospital exposure.10,11 In recent years, there have been 2 studies12,13 showing an increasing CDI trend in children. However, little is known about the trend in severity of CDI in children or about the effect of CDI on children.
There are several known risk factors for the development of CDI, including medication administration (ie, antibiotics, acid suppression medication, and chemotherapy).14- 17 In adults, conditions such as inflammatory bowel disease (IBD) and procedures such as human stem cell transplantation are associated with CDI.18,19 Children with solid tumors, children who underwent organ transplantation, and children with IBD are also at increased risk.20- 22 Although these studies have examined the relationship between CDI and subgroups of patients, there have been no broad investigations of the clinical and demographic risk factors for CDI in children. Using a large national database, we sought to identify trends in CDI and the risk factors associated with CDI, and to evaluate the effect of CDI on hospitalized children.
Data were obtained from the triennial Healthcare Cost and Utilization Project Kids' Inpatient Database (HCUP-KID), sponsored by the Agency for Healthcare Research and Quality. HCUP-KID is the first nationwide inpatient database devoted to children in the United States. It consists of a stratified random sample of 5 754 377 inpatient discharges during the 4 periods of our study: 1997, 2000, 2003, and 2006. Included are discharge data from 22 to 38 states (depending on the year), and for 2006, it represented an estimated 88.8% of all pediatric inpatient visits in the United States. Each record in the database represents a primary hospital discharge and includes up to 14 other diagnostic-related group codes and up to 15 current procedural terminology codes, based on the International Classification of Diseases, Ninth Revision (ICD-9). HCUP-KID assigns an individual-level population weight that allows an estimation of national case rates and trends.23 All children younger than 1 year were excluded because of uncertainty about the true morbidity of C difficile in infants. There is substantial evidence to support C difficile as both normal commensal flora and nonpathogenic in infants younger than 1 year.7,24- 31
The main predictor was the presence or absence of a diagnosis of CDI on an inpatient discharge record. Case selection was performed by searching the database for the diagnostic-related group code 008.45. This is the only code devoted to CDI, and it has been previously validated as an accurate determinant of actual disease.32,33 This code includes 2 descriptions: intestinal infection due to C difficile and pseudomembranous colitis. Demographic data include race, sex, age, payer type, geographic region, and geographic location (urban vs rural) of the hospital. HCUP-KID relies on individual states to provide race-specific data; therefore, racial categorization varies from state to state. To simplify the analysis, race was grouped into white, black, Hispanic, and other based on race classification provided in the HCUP-KID. Payer type was grouped into private insurance or health maintenance organization (HMO), Medicaid/Medicare, and other (no pay, self-pay, or any other). The hospital region was designated Northeast, Midwest, South, or West. A time variable was included to indicate the year the data was collected to test for trends over time.
The main outcomes studied were death rate, colectomy rate, length of hospital stay, and hospitalization charges. Death rate was defined as the number of patients who died per the number of hospital discharges with the diagnostic code for CDI. Colectomy was defined by the following ICD-9 current procedural terminology codes: 45.8 (total intra-abdominal colectomy), 45.7 (partial excision of large intestine), 45.71 (multiple segmental resection of large intestine), 45.72 (cecectomy), 45.73 (right hemicolectomy), 45.74 (resection of transverse colon), 45.75 (left hemicolectomy), 45.76 (sigmoidectomy), and 45.79 (other partial excision of large intestine). Length of hospital stay was evaluated as the number of days hospitalized. Charges were adjusted to 2009 dollars using the consumer price index for inpatient hospital services.34
ICD-9 diagnostic codes for additional comorbid diagnoses in patients who received a diagnosis of CDI were ranked in order of frequency. The most common 80% of codes were examined, and they corresponded to any code for comorbidity that was found at a frequency of 25 times or more. It was felt that diagnoses that occurred less frequently were unlikely to be risk factors for disease. A total of 205 codes were then categorized into clinically similar groups, and a logistic regression model was generated to establish the risk associated with other comorbid conditions. These categories were bacterial infections, fungal infections, cystic fibrosis, neoplastic disease, hematologic disorders, solid organ transplant, hematopoietic stem cell transplantation, malnutrition, IBD, appendicitis, renal disease, liver disease, human immunodeficiency virus, cardiac disease, pancreatitis, gastrostomy tube status, lupus erythematosus, gastroesophageal reflux, and asthma.
Categorical data were summarized as frequencies and percentages, and continuous data were summarized as mean values with standard errors; when data were not normally distributed, they were summarized as median values (interquartile range). Comparisons between groups were conducted using the χ2 test and the t test. Analysis for trend during the 4 periods of our study was performed using multivariate logistic regression, the Cochrane-Armitage test for trend, and analysis of variance for trend. All summary descriptive results and all logistic regression modeling incorporated the HCUP-KID–provided population weights to produce national estimates and appropriate standard errors for significance testing. SAS Enterprise Guide software version 4.2 (SAS Institute, Cary, North Carolina) was used for all statistical analysis.23
To assess the risk factors for CDI among pediatric inpatients, a multivariable logistic regression was fit with the presence or absence of CDI as the dichotomous outcome variable. Covariates were chosen on the basis of theoretical association with CDI. These included demographics (age, sex, race, hospital region, urban vs rural location of the hospital, and payer type), the comorbid conditions identified for this study, and the year of data compilation. Model diagnostics for multicollinearity were assessed using Spearman correlations as well as tolerance and variance inflation factor statistics.
To account for unobservable confounders that were likely to be associated with both CDI and the severity outcomes under study, an analysis including high-dimensional propensity scores was performed. High-dimensional propensity scores were generated by regression analysis of patients with CDI based on their demographics and comorbid conditions, as previously mentioned, and also on the presence or absence of 91 of the most severe pediatric conditions associated with death.35 The inclusion of these additional 91 covariates will either account for specific associations with CDI or collectively serve as proxies for unobserved confounding variables.36 Patients with an indication of CDI (cases) were matched by high-dimensional propensity score, using a greedy matching algorithm, to patients who did not have CDI (controls), with a 1-to-5 matching ratio.37
We used multiple logistic regression models to assess the effect of CDI on the likelihood of death, colectomy, greater than median length of hospital stay, and greater than median hospitalization charges. The models were risk-adjusted for demographic and comorbid condition variables and for the high-dimensional propensity score. Odds ratios (ORs) and 95% confidence intervals (CIs) are presented to identify the strength and significance of CDI and other covariates on the likelihood of cost and severity outcomes.
The total weighted number of pediatric inpatients discharged from the hospital during the 4 years of our study was 10 474 454; of these patients, 21 274 (0.2%) were identified as receiving a diagnosis of CDI. Demographic data by year are summarized in Table 1. The mean (SEM) age of the patients with CDI was 9.50 (0.07) years, and the mean (SEM) age of the other patients was 11.98 (0.01) years (P < .001). There were significant differences in sex (P < .001), race (P < .001), region (P < .02), urban location (P < .001), and payer type (P < .001) between children with CDI and all other children.
There was a large and significant increase in the national trend of cases of CDI in hospitalized children (Figure). The average annual percentage increase of CDI was 14.9% (P < .001). There was not, however, a significant trend in the 4 variables used as surrogates for disease severity (Table 2). The median length of hospital stay remained steady for CDI patients during the 4 periods of our study, at 5 to 6 days (P = .09). Hospitalization charges increased but not significantly, with a mean annual percentage increase of 4.4% for children with CDI vs 2.3% for those without CDI (P = .06). There were also no significant trends in death rate (P = .17) or colectomy rate (P = .09) among patients diagnosed with CDI.
Trend in Clostridium difficile infection (CDI) in hospitalized children.
Comparing data from all 4 years, CDI discharges vs all other hospital discharges showed the significant effect of the disease (Table 2). In multivariate logistic regression, including all comorbidity risk factors of CDI, demographic risk factors of CDI, and the high-dimensional propensity scores, children with CDI had an OR of 1.20 (95% CI, 1.01-1.43) for death, an OR of 1.36 (95% CI, 1.04-1.79) for colectomy, an OR of 4.34 (95% CI, 3.97-4.83) for length of hospital stay longer than the median value, and an OR of 2.12 (95% CI, 1.98-2.26) for hospitalization charges higher than the median value (Table 3).
Children with CDI had a higher median (interquartile range) number of diagnoses than those without (6 [3-9] vs 3 [2-4]; P < .001). In evaluating these comorbidities and other risk factors, we found that there were 19 comorbid categories, 5 demographic variables, and the time (year) included in the multivariate model. The comorbid diagnosis most highly associated with CDI was IBD (OR, 11.42; 95% CI, 10.17-12.83). The ORs were also high for several other comorbid conditions with primary or secondary immunodeficiency or likely exposure to antibiotics (Table 4). The risk of CDI was lower for blacks, Hispanics, and other races compared with whites. The risk of CDI was highest in the West, lowest in the South, and equal in the Midwest compared with the Northeast. The risk was higher in urban hospitals than in rural hospitals. The risk was lower in Medicaid/Medicare patients and other payer types (self-pay or no pay) than in patients with private insurance or HMO.
Among hospitalized children in the United States, we identified a significant effect of CDI, with an increasing incidence and several comorbid and demographic risk factors. There was a significant risk and effect of disease associated with the diagnosis of CDI when it was compared with death rate, colectomy rate, length of hospital stay, and hospitalization charges. We did not find an increase in the severity of CDI over time.
Children with CDI had a greater likelihood of death, colectomy, longer length of hospital stay, and higher hospitalization charges than those without CDI. All of these findings remained statistically significant even after controlling for CDI comorbid conditions associated with the severity outcomes, patient-level demographic variables, and a high-dimensional propensity score associated with acquiring CDI. The uncontrolled death rate, length of hospital stay, and hospitalization charges associated with CDI in children mirror what is seen in adults with CDI.2,38,39 However, the rate of colectomy with the diagnosis of CDI was 10 times less than that reported in adults.2
There was an observed increase in the rate of pediatric CDI during the 4 periods of our study. This increase is consistent with previously published trends in adults and children.2,12,40 Zilberberg et al,13 using an online version of the HCUP-KID, showed that there was an unadjusted increase in CDI in children. Our evaluation of trend confirmed these descriptive findings while controlling for potential confounders in our logistic regression model (Table 4).
One of the primary risk factors for the development of CDI is antibiotic administration. An upward outpatient trend in antibiotic use could explain the inpatient upward trend in CDI. There is evidence, however, of a downward trend in antibiotic administration in the United States during the same period.41 Another, more likely explanation is that there is a more widespread dissemination of a more virulent stain of C difficile, such as the restriction endonuclease analysis group BI, NAP1, chain reaction ribotype 027 (BI/NAP1/027). The increase in the number of cases and their severity in adult patients have been partially attributed to this strain or to an increase in hospitalizations of persons with comorbid diagnoses who are most at risk for CDI.42,43 There may also be increasing awareness among health care providers, leading to increased testing in symptomatic patients.
There was no significant trend in the severity of CDI in children as measured by length of hospital stay, hospitalization charges, death rate, colectomy rate, and death. The lack of increase in both rates of colectomy and death associated with CDI is in agreement with a previous report in children and is in contrast to findings in adult studies.2,12 We did observe an increase in the consumer price index–adjusted charges over time; however, this increase was not statistically significant (P = .06).
Inflammatory bowel disease was the most highly associated comorbid diagnosis with CDI. Children with IBD were 11 times more likely to also have a diagnosis of CDI than children without IBD. This supports previous reports of both adults and children.19,44,45 Comorbidity is also a known risk factor for CDI in adults, and we have shown this in children.46 The additional diagnostic groups identified with an increased risk of CDI were similar for known risk factors previously reported, such as hematopoietic stem cell transplantation, solid organ transplantation, and solid tumors.18,20,21,47 These diagnostic categories, in addition to the others reported in our study, reinforce the known risk of immune suppression or antibiotic exposure.48- 50 Increased awareness of groups at highest risk could decrease adverse outcomes, such as death and colectomy, through early testing of symptomatic patients and earlier appropriate treatment.
The finding that CDI is more common in male children is in agreement with previous reports in pediatrics.12 However, this sex discrepancy is reversed in adult hospitalized patients.2 Patients with CDI were significantly younger than patients without CDI. This could be explained by increased antibiotic pressure in younger children, bias to test for CDI in younger patients, or a true susceptibility to disease at a younger age. In adults, it is the elderly who are at the greatest risk of disease; thus, CDI may be a disease affecting the extremes of age, with adolescents and younger adults being relatively spared. We found a higher risk of CDI in the West and a lower risk in the South. A previous national study of primarily adult cases of CDI showed the highest incidence to be in the Northeast.51 In the overall analysis, there was a decreased risk of CDI associated with Medicaid or other payer types (self-pay and no pay). This may represent a decreased risk of disease in these patients because of a lower exposure to antibiotics for routine infections (eg, acute otitis media) compared with those with private insurance.
In evaluating the severity markers, we identified that female patients had the same risk of mortality; however, they had a longer length of hospital stay and higher hospitalization charges. Blacks had an increased risk of death, a longer length of hospital stay, and higher hospitalization charges. We were not able to examine the full scope of these findings; however, further research into the disparities of care may be justified in this area. There was an increased risk of all severity markers at urban hospitals vs rural hospitals, which, despite controlling for many factors, may still be due to referral bias not captured in the analysis.
The HCUP-KID provides an excellent national representative sample of pediatric hospitalizations and can be used to extrapolate nationwide estimates and trends. To our knowledge, the sample size allowed us to conduct the largest and broadest reported study of CDI in children. Use of HCUP-KID does, however, have several limitations. Our definition of CDI was based entirely on ICD-9 coding. As in all studies using diagnostic codes, there is concern for the paucity of clinical information. We were unable to verify the specific strain of CDI, the medical treatments given, or the location of an acquired infection (outpatient vs nosocomial). This is particularly important because community-associated disease is increasingly recognized as another route of infection and because not all cases are associated with exposure to health care facilities or antimicrobial treatment. Misclassification and reporting bias are both potential limitations. Cdifficile is a common commensal organism found in infants.24- 30 It is unclear at what age an infant must reach before C difficile becomes pathologic, but it is likely after the first year of life.7,31 We excluded children younger than 1 year in an effort to avoid misclassification bias. Previous studies in both adults and children have validated the ICD-9 code for CDI as representing true disease.32,33 There is more recent evidence that C difficile may also exist in a carrier state (eg, in patients with IBD).52 Because the HCUP-KID reports hospital discharges (not individual patients), it was not possible to distinguish patients readmitted to the hospital for the same diagnosis.
The population-based data in our study provide additional evidence that CDI cases have a significant effect on the pediatric population. Our study supports previous reports that CDI is increasing among hospitalized children and provides a background for understanding changing trends and risk factors of CDI in children. Increasing awareness of these risk factors and of an upward trend in hospitalized children with CDI is the first step in controlling this important infection.
Correspondence: Cade M. Nylund, MD, Department of Pediatrics, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Rd, Bethesda, MD 20814 (email@example.com).
Accepted for Publication: October 7, 2010.
Published Online: January 3, 2011. doi:10.1001/archpediatrics.2010.282
Author Contributions: Drs. Nylund and Goudie had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Nylund, Goudie, Garza, and Cohen. Acquisition of data: Goudie. Analysis and interpretation of data: Nylund, Goudie, Garza, Fairbrother, and Cohen. Drafting of the manuscript: Nylund and Goudie. Critical revision of the manuscript for important intellectual content: Nylund, Goudie, Garza, Fairbrother, and Cohen. Statistical analysis: Goudie. Administrative, technical, and material support: Goudie, Garza, and Cohen. Study supervision: Fairbrother and Cohen.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant HS 016957 from the Agency for Healthcare Research and Quality (Dr Goudie).
Disclaimer: Dr Nylund is employed by the US Air Force. The views expressed in this article are those of the authors and do not reflect the official policy or position of the US Air Force, the Department of Defense, or the US government.
Previous Presentation: The results presented were presented at the proceedings of Digestive Disease Week; May 3, 2010; New Orleans, Louisiana.
Additional Contributions: Pamela J. Schoettker, MS, Medical Writer, Division of Health Policy and Clinical Effectiveness, Cincinnati Children's Hospital Medical Center, provided editorial assistance.
Country-Specific Mortality and Growth Failure in Infancy and Yound Children and Association With Material Stature
Use interactive graphics and maps to view and sort country-specific infant and early dhildhood mortality and growth failure data and their association with maternal
Thank you for submitting a comment on this article. It will be reviewed by JAMA Pediatrics editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Register and get free email Table of Contents alerts, saved searches, PowerPoint downloads, CME quizzes, and more
Subscribe for full-text access to content from 1998 forward and a host of useful features
Activate your current subscription (AMA members and current subscribers)
Purchase Online Access to this article for 24 hours
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 4
Customize your page view by dragging & repositioning the boxes below.
and access these and other features:
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.