Diarrhea Etiology in a Children's Hospital Emergency Department: A Prospective Cohort Study

Abstract

Background. We evaluated the frequency of recovery of pathogens from children with diarrhea who presented to a pediatric emergency department and characterized the associated illnesses, to develop guidelines for performing a bacterial enteric culture.

Methods. We conducted a prospective cohort study of all patients with diarrhea who presented to a large regional pediatric emergency department during the period from November 1998 through October 2001. A thorough microbiologic evaluation was performed on stool specimens, and the findings were correlated with case, physician, and laboratory data.

Results. A total of 1626 stool specimens were studied to detect diarrheagenic bacteria and, if there was a sufficient amount of stool, Clostridium difficile toxin (688 specimens), parasites (656 specimens), and viruses (417 specimens). One hundred seventy-six (47%) of 372 specimens that underwent complete testing yielded a bacterial pathogen (Shiga toxin-producing Escherichia coli, 39 specimens [of which 28 were serotype O157:H7]; Salmonella species, 39; Campylobacter species, 25; Shigella species, 14; and Yersinia enterocolitica, 2), a viral pathogen (rotavirus, 85 specimens; astrovirus, 27; adenovirus, 18; or rotavirus and astrovirus, 8), a diarrheagenic parasite (5 specimens); or C. difficile toxin (46 specimens). Samples from 2 patients yielded both bacterial and viral pathogens. A model to identify predictors of bacterial infection found that international travel, fever, and the passing of >10 stools in the prior 24 h were associated with the presence of a bacterial pathogen. Physician judgment regarding the need to perform a stool culture was almost as accurate as the model in predicting bacterial pathogens.

Conclusions. Nearly one-half of the patients who presented to the emergency department with diarrhea had a definite or plausible pathogen in their stool specimens. We were unable to develop a model that was substantially better than physician judgment in identifying patients for whom bacterial culture would yield positive results. The unexpectedly high rate of C. difficile toxin warrants further examination.

The etiologies of diarrhea in North American children have not been well studied in recent years. Furthermore, guidelines about bacterial culture [1] have not been validated with regard to recently emerging agents, such as Shiga toxin-producing Escherichia coli (STEC).

The Children's Hospital and Regional Medical Center (CHRMC) emergency department in Seattle, Washington, serves a demographically varied urban and suburban population with a broad range of illness severity. Our objectives were to determine the etiologies of diarrhea in children presenting to this emergency department and to model the variables associated with the presence of bacterial enteric pathogens. The overall goal was to provide data to clinicians, to guide them in the rational use of cultures to detect bacterial pathogens. Several previous publications have resulted from this study [2–4].

Methods

All patients who presented with diarrhea to the CHRMC emergency department during the period from November 1998 through October 2001 were eligible for enrollment in the study. Parents were given an information sheet about the study and its voluntary nature, and if they agreed to enrollment, they were instructed to complete a questionnaire in English, Russian, Spanish, Somali, or Vietnamese that addressed their history and demographic characteristics. If the patient could not provide a stool specimen during the visit, a swab specimen from the rectum was obtained, if the family consented. The physician who cared for the patient in the emergency department was also asked to complete a form that included vital signs and medical care given and to designate whether he or she would have ordered any evaluations on the patients had this study not been underway that day. The procedures followed were in accordance with the ethical standards of the CHRMC Institutional Review Board and the Helsinki Declaration of 1975, as revised in 1983.

In the laboratory, stool specimens were evaluated for visible blood and for RBCs and WBCs by microscopic evaluation by the guiac test for occult blood, with specimens plated on sheep's blood, MacConkey, sorbitol-MacConkey, Hektoen, and Salmonella-Shigella-, Campylobacter-, and Yersinia-selective (Prepared Media Laboratories) agars and inoculated into Selenite F (BBL; Becton Dickinson) and MacConkey (Binax) broths. Campylobacter plates were incubated microaerophilically at 42°C; all other media were incubated at 35°C. These techniques are adequate to isolate Aeromonas species, Campylobacter species, E. coli O157:H7, Pleisiomonas shigelloides, and Salmonella, Shigella, Vibrio, and Yersinia species. The MacConkey broth was tested after overnight incubation by EIA (Meridian Biosciences) for the presence of Shiga toxin. O157:H7 and non-O157:H7 STEC were identified on sorbitol-MacConkey agar and by the PCR of isolated colonies, respectively [3].

If stool sample volumes were sufficient, we sought to detect Clostridium difficile toxin using a cytotoxicity assay with cultured human diploid fibroblasts, with cytotoxicity confirmed by neutralization with Clostridium sordellii antitoxin; trichrome stains and formalin/ethyl acetate sedimentation were used for detection of parasites, and fluorescence antibody testing was used for detection of Giardia and Cryptosporidium species (Techlab). For frozen stool specimens, we used Rotaclone (Meridian Biosciences), Adenoclone (Meridian Biosciences), and astrovirus (IDEIA Astrovirus EIA kit, DAKO) EIAs for detection of rotavirus, adenoviruses, and astrovirus, respectively.

We evaluated associations between bona fide or candidate diarrheagenic bacteria, viruses, and parasites and historical and clinical features in univariate analyses. We determined relative risks (RRs) and their 95% CIs for univariate analysis. We used Pearson's χ² test and Fisher's exact test to assess the significance of differences between groups (for categorical data), and we used the t test for independent groups for evaluation of continuous data (assuming unequal variance). P values <.05 were considered to be statistically significant.

We also performed a predictive logistic regression analysis on the 1238 patients for whom no gross blood was reported by the patient's parents, to construct a model that would potentially help clinicians decide which patients were likely to carry bacterial enteric pathogens on the basis of history and examination findings. Results are reported as ORs with 95% CIs. A history of blood in stool was not included in the model, because we assumed that clinicians would generally culture stool samples from patients with bloody diarrhea. Variables that were modeled included patient age; use of antibiotics in the previous 30 days; travel outside of the United States in the previous 30 days; duration of diarrhea (in days); number of stools passed in the 24 h before the visit; history of abdominal pain, vomiting, and fever; and fever and/or abdominal tenderness noted during examination. Because this was an exploratory exercise, we retained variables in the model at a significance level of .10.

Results

Of the 4799 patients who were discharged from the CHRMC emergency department with diarrhea, bloody diarrhea, or gastroenteritis during the study period, 1626 (33.9%) agreed to enroll in the study and were able to provide a specimen, which, at a minimum, underwent bacterial culture. Specimens were collected by cup (for 699 patients), diaper (for 638 patients), swab (for 252 patients), or unknown methodology (for 37 patients). The mean age of all patients enrolled was 2.5 years (median age, 1.3 years); 57% of patients were male. of the patients enrolled, 53% were white, 11% were African American, 11% were Asian, 1% were American Indian, 16% were of another or mixed race, and 8% did not respond to this question; 21% were self-identified as Hispanic. Only 4% of all enrollees had recently traveled outside of the United States. The subjects had passed an average of 8 stools in the 24 h before presentation (median, 6 stools; range, 1–60 stools), and the median duration of diarrhea before presentation was 3 days (range, <1–360 days). Fifteen percent of patients reported having blood in their stool.

The median age of the patients who were infected with a bacterial enteric pathogen was 4.7 years, compared with 1.2 years for patients with no bacterial pathogen detected. The sex, race, ethnicity, and number of days of diarrhea before presentation were similar for children with positive bacterial culture results and those with negative bacterial culture results. Eleven percent of enrollees with a positive bacterial culture result had traveled recently outside of the United States, compared with 4% of those with negative culture results (P < .001). The median numbers of stools passed in the 24 h before evaluation were 10 and 6 for patients with positive and patients with negative culture results, respectively (P < .001). Finally, 62% of patients with positive culture results reported having blood in their stool, compared with 12% of those with negative culture results (P < .001) (table 1).

Table 1

Demographic and illness characteristics in a study of diarrhea etiology from a children's hospital emergency department.

Organisms generally accepted to be pathogens (i.e., “presumptive pathogens”) were recovered from the stool samples from 261 patients. One hundred eighteen (7.3%) of 1626 stool samples tested contained bacterial pathogens, 138 (33%) of 417 stool samples tested contained viral pathogens, and 7 (1.1%) of 656 stool samples tested contained parasitic pathogens; 2 patients were infected with both bacterial and viral pathogens. In addition, candidate pathogens (C. difficile and Blastocystis hominis) were identified in 53 patients' stool samples (figure 1). Overall, 176 (47%) of 372 children whose stool specimens underwent a complete battery of tests were found to be carrying at least 1 presumptive pathogen (table 2).

Figure 1

Results of tests for diarrheagenic pathogens. Bona fide and candidate diarrheagenic pathogens were recovered, as indicated. Results are expressed as number of pathogen-positive subjects divided by the number tested. B. hominis, Blastocystis hominis; C. difficile, Clostridium difficile; N, number of subjects; n, number of specimens. *One hundred eighteen stool specimens tested positive for bacterial pathogens, and 119 bacterial pathogens were detected.

Table 2

Pathogens and candidate pathogens identified in 372 stool specimens that were subjected to a full battery of testing.

Stool specimens collected in the summer were more likely to yield bacterial pathogens, and those collected in the winter and spring were more likely to yield viral pathogens (table 3). Leukocytes were present in 48%, 8%, and 29% of specimens containing bacterial, viral, and parasitic pathogens, respectively. Results varied by age. Viruses were identified in 55% of stool samples obtained from patients aged <3 years and in 23% of stool samples obtained from those aged ⩾3 years (P < .001). Diarrheagenic bacterial pathogens were found in 22% of stool samples obtained from patients aged ⩾3 years and in stool samples obtained from 4% of patients aged <3 years (P < .001). There was no association between age and detection of a diarrheagenic parasite. Culture results also varied by collection method. A bacterial pathogen was identified in 20% of specimens obtained using a cup, in 7% of those obtained using a diaper, and in 1% of swab specimens (P < .001), although these results could reflect, in part, the age of the patient. Overall, a bacterial pathogen was recovered from stool samples obtained from 7.3% of all patients tested, with STEC and Salmonella species being the most common bacterial pathogens (table 4). E. coli O157:H7 was the predominant STEC, as previously reported in this population [3] and in another Seattle-based population [5].

Table 3

Seasonality of pathogen recovery in a study of diarrhea etiology from a children's hospital emergency department.

Table 4

Bacterial pathogens recovered in a study of diarrhea etiology from a children's hospital emergency department.

In univariate analyses, factors associated with the presence of a bacterial pathogen on stool culture included travel outside the United States in the previous 30 days (RR, 2.8; 95% CI, 1.7–4.8), blood in the stool reported by the parent (RR, 7.4; 95% CI, 5.0–10.9), the passing of >10 stools in the previous 24 h (RR, 1.1; 95% CI, 1.1–1.2), history of abdominal pain (RR, 2.1; 95% CI, 1.2–3.6), and abdominal tenderness (RR, 3.8; 95% CI, 2.4–5.7). Factors associated with the absence of a bacterial pathogen included antibiotic use in the previous 30 days (RR, 0.3; 95% CI, 0.3–0.7), duration of diarrhea of >10 days (RR, 0.3; 95% CI, 0.1–0.8), and vomiting (RR, 0.5; 95% CI, 0.3–0.7).

A viral pathogen was identified in 138 (33.1%) of the 417 patients tested (table 5). Eight stool specimens (1.9%) yielded both rotavirus and astrovirus (table 5). Factors associated with the presence of a viral pathogen included antibiotic use in the previous 30 days (RR, 1.8; 95% CI, 1.4–2.4), vomiting (RR, 4.7; 95% CI, 2.7–8.2), and fever (RR, 1.6; 95% CI, 1.1–2.3). Parental report of blood in the stool was negatively associated with viral infection (RR, 0.6; 95% CI, 0.5–0.7).

Table 5

Viral pathogens identified in a study of diarrhea etiology from a children's hospital emergency department.

A diarrheagenic parasite was identified in only 1.1% of all stool specimens that were tested for such agents (table 6). Giardia lamblia was detected in 4 stool specimens by microscopic evaluation; 3 of these infections were also detected by antigen testing. Both Cryptosporidium parvum infections were detected by microscopic evaluation and antigen testing. The only significant association with parasitic infection was travel outside of the United States in the previous 30 days (RR, 8.5; 95% CI, 1.7–42.0), but only 2 of 7 patients infected with a diarrheagenic parasite reported this particular risk factor during that interval.

Table 6

Diarrheagenic parasites identified in a study of diarrhea etiology from a children's hospital emergency department.

C. difficile toxin and B. hominis were found in 6.7% and 1.1% of stool samples tested, respectively. For children aged ⩾3 years whose families answered this question, there was a strong association between recent use of antibiotics and presence of C. difficile toxin. Among patients aged 0–3 years who were tested for C. difficile and who had recently used antibiotics, test results were positive for 12 (9%) of 128, compared with 15 (5%) of 310 patients who had not recently used antibiotics (P = .1). of patients aged ⩾3 years who were tested for C. difficile and who had recently used antibiotics, test results were positive for 13 (23%) of 56, compared with 3 (2%) of 172 patients without exposure to antibiotics (P < .0001).

Multivariate analysis to identify factors that might help predict the presence of a bacterial pathogen on stool culture demonstrated the following positive associations: older age (OR, 1.2; 95% CI, 1.2–1.3), travel out of the United States (OR, 3.7; 95% CI, 1.7–8.0), fever (OR, 2.3; 95% CI, 1.3–4.2), and passing of >10 stools in the previous 24 h (OR, 4.5; 95% CI, 2.7–7.4). Conversely, vomiting (OR, 0.4; 95% CI, 0.2–0.6), antibiotic use in the previous 30 days (OR, 0.4; 95% CI, 0.2–0.8), and duration of diarrhea of >10 days (OR, 0.3; 95% CI, 0.1–0.9) were negatively associated with the identification of a bacterial pathogen. In this study, for a patient who traveled outside the country, had fever, produced >10 stools in the previous 24 h, and had diarrhea of ⩽10 days' duration but who had not been vomiting or had not used antibiotics in the previous 30 days, the probability of having a bacterial pathogen on stool culture was 85%. A patient who had no international travel, who had no fever, and who had passed ⩽10 stools in the previous 24 h but who had been vomiting, had used antibiotics in the previous 30 days, and had diarrhea of >10 days' duration had a <1% probability of having a bacterial pathogen detected by stool culture.

Physician judgment was almost as predictive as the logistic model for detection of stool samples that contained a bacterial pathogen. of the 79 patients with a bacterial pathogen, physicians would have requested a bacterial culture for 60 (76%). of the 1067 patients in whom a bacterial pathogen was not isolated, the treating physician would have ordered a stool culture for only 372 (35%). Thus, physician judgment has a sensitivity of 76% and a specificity of 65%.

Discussion

This study provides information regarding the etiologies and diagnostic strategies appropriate for children presenting with diarrhea to an urban pediatric emergency department in a resource-rich country, and it demonstrates that intensive analysis of stool specimens in such a high-acuity population can identify a larger proportion of pathogens than has been reported previously [1, 6–8]. Specifically, 47% of specimens that underwent complete testing contained a known or presumptive pathogen, even without seeking noroviruses.

STEC—in particular, E. coli O157:H7—and salmonellae were the most common bacterial pathogens isolated. Campylobacter species were less common than in other studies [7]. Although Aeromonas species were found in 4.8% of Swiss children hospitalized with diarrhea [9], this candidate pathogen was not isolated in our study.

Rotavirus was, as expected, the most common pathogen. The appreciable detection rate of astroviruses, as has been reported in Australia [10, 11], Rhode Island [12], and California [13], suggests that this group of viruses is an important (but overlooked) cause of childhood diarrhea.

In one study of outpatient nondysenteric diarrhea in various regions of the United States, 6.8% of patients were found to be infected with G. lamblia [14]. However, the emergency department setting of our study was probably biased against detection of patients infected with agents that cause chronic diarrhea.

The pathogenicity of B. hominis in the patients in whom this agent was detected is not clear, although in one recent report, B. hominis was found to be more common in control subjects than in patients with diarrhea [6]. Future studies should include an appropriate control population, to confirm or refute the role of B. hominis as a cause of diarrheal illnesses in children. A similar situation exists with C. difficile. C. difficile toxin can be found in stool specimens obtained from children without diarrhea at rates equal to those with diarrhea in hospitals [15]. However, the 6.7% positivity rate is high, and the median age of C. difficile-positive patients in this study's population (19 months) is somewhat higher than the median age of asymptomatic children whose stool samples contain this toxin (∼11 months) [15–18]. Our data raise the possibility that the presence of C. difficile is more likely to be associated with antibiotic use in older children, although our study, which lacked a control group of subjects who did not receive antibiotics, did not address the causative association between antibiotic use and diarrhea. Also, our estimation of the frequency of C. difficile infection may be somewhat low, because our study used the cytotoxicity assay, which only detects toxin B; therefore, we would not have detected toxin A-positive, toxin B-negative strains, which can be present in children's stools [19].

We wish to make some comments regarding the use of guidelines or modeling in the decision whether to obtain stool samples for culture. Although the logistic model indicating presence of a bacterial pathogen was only modestly better than physicians' judgment about whether to obtain stool samples from a particular patient for culture, it should be noted that the high-acuity setting of this emergency department study may have enabled physicians to make judgments regarding culture based on more experience, and that algorithms may have a better role in defining inclusion and exclusion criteria in other ambulatory sites. Also, in view of our experience with patients with STEC infection [3, 20], who are rarely febrile when they present for medical attention, we caution against using the absence of fever as a reason to forego a stool culture. Fecal leukocytes are more common in stool specimens that contain bacterial pathogens, but 52% of stool specimens that test positive for a bacterial pathogen did not contain fecal leukocytes. Although the presence of fecal leukocytes indicates the need for a culture for bacterial enteric pathogens, the absence of fecal leukocytes should not deter clinicians from performing a culture.

Although this etiologic study involves among the largest cohorts and most intensive analyses of diarrhea in children to have been recently reported, it does have limitations. One limitation is that we were only able to enroll one-third of eligible patients. This speaks to the difficulty in obtaining stool specimens from patients with diarrhea who are on site for a short period of time; most patients simply did not provide a stool specimen during their emergency department stay and declined the offer to obtain a diagnostic rectal swab. Other limitations include the lack of control subjects, our inability to explain the etiology of many cases of diarrhea, and our lack of evaluation for other viruses and for diarrheagenic E. coli other than STEC, such as enteroaggregative, enterotoxigenic, and enteropathogenic E. coli, for which there are no commercially available tests. Indeed, noroviruses, which we did not attempt to detect, have been reported as pathogens in other studies in which their presence was sought [6, 21]. In addition, our study population may differ from the population of other studies because of geographic variations in prevalences of pathogens. For these reasons, future investigations of diarrhea etiology should include appropriate control populations to assess the frequencies with which pathogens and candidate pathogens are found in asymptomatic subjects, and they should also incorporate techniques to identify pathogens not sought in the present study.

Cohen et al. [22] provide complementary as well as contrasting data. These investigators, from Cincinnati, Ohio, found that rotavirus was the most common etiologic agent in childhood diarrhea, as did we, but they identified classic bacterial pathogens considerably less frequently (2.1%). They sought neither C. difficile toxin nor viruses other than rotavirus; therefore, it is not possible to compare our rates of recovery of these agents with theirs. Conversely, they made a case for the pathogenicity of enteroaggregative, diffuse adherent, and atypical enteropathogenic E. coli in their study population [23]. These agents were not addressed in our study, so their presence cannot be excluded.

Interestingly, the spectrum of etiologic agents that we identified differed from the spectrum identified by Denno et al. [4] in other ambulatory (non-emergency department) children in Seattle; in our study, we identified proportionately more cases involving STEC and astrovirus. Clearly, there exist geographic differences in the spectrum of diarrheal agents, and the range of pathogens and candidate pathogens identified is critically dependent on the employed technology. Furthermore, facility use behavior could also be influenced by the severity of illness, which varies according to etiologic agent, thereby skewing the microbial results on the basis of the site. Similarly, our study of an emergency department-based population identified a considerably greater number of plausible pathogens than a home-based nationwide surveillance study by Vernacchio et al. [24]. However, taken together, these studies and the data we present demonstrate the need for—and opportunities in—expanded diarrhea etiology investigations in North America.

In summary, a definite or plausible causative agent was detected in almost one-half of the patients in this study who underwent a complete battery of testing, even though the panel of tests that we used did not address every conceivable pathogen. We were unable to develop a model substantially better than physician judgment to identify patients for whom a stool culture would yield positive results. The unexpectedly high rate of C. difficile toxin detection in this study suggests an area for future examination. The etiologies of unexplained childhood diarrhea warrant further investigation.

Acknowledgments

We thank Elizabeth Wolf and Jennifer Falkenhagen for assistance in manuscript preparation; Evangeline Sowers and Katherine Greene for serotyping efforts; the Children's Hospital and Regional Medical Center Microbiology Laboratory staff for the high standards of testing used in this study; Karen Wygant and the Emergency Department staff for their assistance with this project; and Drs. Joseph Bresee, Vance Dietz, Thomas Navin, Caryn Byrn, Patricia Griffin, and Robert Tauxe, and Mr. Greg Jones at the Centers for Disease Control and Prevention for their input and expertise.

Financial support. Centers for Disease Control and Prevention (cooperative agreement CCU015040).

Potential conflicts of interest. All authors: no conflicts