Differential yield of pathogens from stool testing of nosocomial versus community-acquired paediatric diarrhea

Presented in part at the 36th Interscience Conference of Antimicrobial Agents and Chemotherapy (Paper J153), New Orleans, September 15 to 18, 1996 Departments of Microbiology and Infectious Diseases, and Pediatrics, University of Calgary, Alberta Children’s Hospital, Calgary, Alberta Correspondence and reprints: Dr H Dele Davies, Child Health Research Unit, Room 2271, Alberta Children’s Hospital, 1820 Richmond Road SW, Calgary, Alberta T2T 5C7. Telephone 403-229-7813, fax 403-541-7508, e-mail dele.davies@crha-health.ab.ca Received for publication November 16, 1998. Accepted March 2, 1999 S Deorari, A McConnell, K-K Tan, et al. Differential yield of pathogens from stool testing of nosocomial versus community-acquired paediatric diarrhea. Can J Infect Dis 1999;10(6):421-428.

D iarrheal diseases are a major cause of morbidity and hos- pitalization in North American children, and a leading cause of morbidity and mortality in children worldwide.It has been estimated that 16.5 million American children younger than five years of age have between 21 and 37 million episodes of diarrhea annually, and approximately 10.6% of hospitalizations in this age group are for diarrhea (1).Gastrointestinal infections account for 16.8% to 20% of hospital-acquired disease (2,3), and it was estimated that the mean cost per nosocomial infection was US$836 (1977 dollars) for diagnostic and therapeutic measures (4).Knowledge of etiological agents of infectious diarrhea at a particular institution may help to determine focused empirical therapy and to choose the appropriate stool tests, possibly reducing the cost of hospitalization without affecting patient care.Previous studies in other localities have documented significant site to site differences with respect to the types of pathogens causing paediatric infectious diarrhea (5)(6)(7)(8)(9)(10)(11)(12), but have not clearly differentiated community-acquired diarrhea (CAD) from nosocomial diarrhea (NAD) (5)(6)(7)9,11).A wide range of viral, bacterial and parasitic causes of infectious diarrhea are now recognized.Because of the variety of organisms, it is difficult and expensive for clinical laboratories to do a complete etiological examination of diarrheal stool specimens.Some investigators, mostly studying adult populations, have recently questioned the appropriateness of performing a complete etiological examination on stool specimens for diarrheal diseases (13)(14)(15)(16)(17)(18).The few published reports on stool testing in the paediatric population have either not differentiated CAD from NAD (19)(20)(21), or have excluded viral testing or screening for ova and parasites (17)(18)(19)(20)(21). Furthermore, it was unclear what the practice of screening diarrheal NAD stool specimens was among Canadian paediatric centres.In view of these considerations, we sought to characterize the infectious agents associated with both CAD and NAD at our paediatric institution and the characteristics of diarrhea due to different pathogens, and evaluate the appropriateness of screening for all pathogens in NAD in children.

Patient profile:
The Alberta Children's Hospital is a tertiary regional referral hospital in Calgary that is affiliated with the University of Calgary, Alberta.The hospital serves a population base of 1.2 million children, seeing children referred from southern Alberta, southeastern British Columbia, northern Montana and southwestern Saskatchewan.From April 1, 1993 to March 30, 1995, the health and microbiological records of all children with an admission or discharge diagnosis of diarrhea were reviewed.Using a standardized data collection form, the following information was abstracted from the patient records: age, sex, date of admission, frequency and duration of diarrhea, character of diarrhea, associated symptoms, whether specimens were obtained for different enteropathogens, results of such testing and the outcome of the patient.In addition, in July 1998, a telephone survey of laboratories serving all paediatric hospitals in Canada was conducted using a standard questionnaire to obtain information with regard to their practices for screening for pathogens related to NAD.Definitions: Diarrhea was defined as stools unusually loose or frequent compared with the norm for each child, as perceived by the caregiver (22).A diarrheal episode was defined as occurrence of diarrhea in a child following more than one week of being well (22).NAD was defined as an occurrence of diarrhea in a child after 72 h of hospitalization, while all others were considered community acquired (23).A diarrheal episode was considered to be infectious if a pathogen was isolated from stool.Recent antibiotic use was defined as the use of antibiotics within one month of presentation.An episode was defined as having significant bacteremia if the bacteremia was treated with antibiotics.Stool examination: Stool specimens were collected and transported in Enteric Plus (Dalynn Laboratory Products, Calgary, Alberta) media.The specimens were routinely inoculated onto a number of selective and differential media, including blood agar, MacConkey, Hektoen agar, Yersinia agar (CIN), MacConkey agar containing sorbitol, a campylobacter agar (CAMPY) and selenite broth (SEL-F).All plates and the SEL-F broth tube

422
Can J Infect Dis Vol 10 No 6 November/December 1999 were aerobically incubated at 37°C, except for the CAMPY plate, which was incubated in a microaerophilic atmosphere at 42°C.Enteric pathogenic isolates were screened off the primary plates using triple-sugar iron, urea, and motility/indole/lysine slants.Other routine biochemical tests, serotyping, and a VITEK (bio-Mérieux, St Louis, Missouri) Gram-negative identification card were used to identify each pathogenic stool isolate fully.Clostridium difficile was cultured on cefoxitin cycloserine fructose agar (CCFA ).Stool was tested for C difficile toxin using an enzyme immunoassay (EIA) (Cambridge Biotech, Worcester, Massachusetts).Stool was tested for rotavirus and adenovirus by EIA (Cambridge Biotech).Electron microscopy (EM) for rotavirus was performed on a 50% stool suspension.Stool was cultured for adenovirus and enterovirus on cultures of LLC-monkey kidney cells (LLC-MK) and human carcinoma cells.Viral culture and EM were performed on request and if EIA testing was negative.All stool specimens for ova and parasites were collected and transported in sodium acetyl formalin.Preserved stools were then concentrated using a standard acetate-acetic acid-formalin fixative (24).Two slides were prepared: one slide, made from the washed stool, was stained permanently using a modified iron hematoxyline/Kinyoun stain (25), and another slide from the stool concentrate was examined unstained.Statistical analysis: Differences in group proportions were compared by use of the c 2 test or the Fisher's exact test.Differences in median were tested for significance using the Mann-Whitney U test.All statistical calculations were performed with Statistica version 5.0 (Statsoft Incorporated, Tulsa, Okalahoma).In all analyses, a two-tailed P<0.05 was considered statistically significant.

RESULTS
Epidemiology: During the study period, 12,305 patients were admitted to the hospital.Five hundred and twenty-three episodes of diarrhea were identified; 434 episodes occurred among 399 (3.2% of all discharges) patients with CAD (353 episodes/10,000 discharges/year) and 89 episodes of diarrhea in 81 patients (0.7% of discharges) with NAD (72 episodes/10,000 discharges/year).The median age of children with CAD at the onset of the diarrheal episode was 19.8 months (range 0.13 to 215.1 months), while for NAD episodes the median age was 19.0 months (range 0.17 to 209.9 months, P=1.00).Two hundred and forty-seven (56.9%)CAD episodes occurred in males versus 47 (52.8%)NAD episodes (P=0.48).Rotavirus EIA and bacterial cultures were performed on the majority of diarrheal episodes, while less than 30% were subjected to viral culture, EM, adenovirus antigen testing, and ova and parasite screen (Table 1).NAD episodes were subjected to adenovirus antigen and C difficile toxin testing more often than CAD.Organisms identified: An infective agent was identified in 34.8% of all diarrheal episodes.Among patients with CAD, pathogens were identified in 143 (32.9%) episodes (116 episodes/10,000 discharges) versus 39 (43.8%)NAD episodes (31.7 episodes/10,000 discharges).Among patients who received specific diagnostic tests, pathogens were identified in 143 of 363 (39.4%) of CAD and 39 of 84 (46.4%) of NAD epi-sodes.The majority of the diarrheal episodes in NAD were associated with a single pathogen; only one of 84 (1.2%) samples from NAD episodes tested was associated with two organisms (C difficile and Blastocystis hominis).Similarly, CAD episodes were also predominantly associated with a single pathogen.Multiple enteropathogens were identified in only six of 363 (1.7%) CAD episodes tested: C difficile and adenovirus in one, rotavirus and adenovirus in one, campylobacter and B hominis in one, Escherichia coli O157:H7 and Enterobius vermicularis in one, rotavirus and B hominis in one, and salmonella, aeromonas and rotavirus in one.
Viruses were the organisms most commonly identified in the CAD and NAD specimens tested, accounting for one-third of each group.Bacteria were also frequent, but non-C difficile bacteria were never identified in any of 57 NAD diarrheal episodes tested.Rotavirus and C difficile were the most commonly identified pathogens in both CAD and NAD episodes (Table 2), accounting for one-fourth to one-third of all tested episodes.The majority of rotavirus infections occurred in the winter (P<0.0001),whereas C difficile toxin was identified mainly in the summer, fall and early winter (P<0.001).Screening for ova and parasites were positive in 11 CAD episodes, and the organisms identified were B hominis (n=4), Giardia lamblia (n=3), Cryptosporidium species (n=2), Dientamoeba fragilis (n=1) and Enterobius vermicularis (n=1).Whereas parasites were identified in 8.6% of CAD episodes tested, they were identified in only one NAD episode during which B hominis was found in association with campylobacter infection.Specific testing by diarrheal episode: Rotavirus was detected primarily by EIA antigen screening in both CAD and NAD specimens (Table 2).The use of EM revealed rotavirus in a further 2% of cases.Viral culture was successful in isolating adenovirus in less than 10% of episodes tested, but was of greater value than adenovirus 40     specimens sent from June to December (P<0.001),while the other bacterial pathogens were isolated mainly in July and August (P<0.0001)(Figure 2).Almost one-third of all CAD episodes underwent ova and parasite screening, with less than 10% of the specimens being positive (Table 2).Testing of NAD episodes had a negligible yield.CAD episodes occurring in children older than five years of age were just as likely to be positive for parasites as episodes in younger children (P=0.73).Although 12.5% of parasite-positive episodes occurred during the period June to August, there was no season in which parasites were more likely to be identified (P=0.47).Clinical information: Viral CAD episodes were significantly associated with day care attendance, vomiting, watery diarrhea and abdominal tenderness (Table 3).CAD episodes in which bacterial pathogens were identified were linked with nausea, malaise, loss of appetite, blood or mucous in the stool, eating fast food recently, other family members with diarrhea and abdominal tenderness (Table 4).Antibiotic use was the only risk factor associated with bacterial NAD cases (11 of 17 bacterial episodes versus 25 of 70 nonbacterial episodes, P<0.02).Compared with episodes in which other bacteria were isolated, C difficile-associated diarrhea was more commonly associated with recent antibiotic use (63.0%versus 20.7%, P<0.001) and chemotherapy (32.8% versus 0%, P<0.001).
In a logistic regression model including vomiting, nausea, headache, watery versus bloody diarrhea, mucousy diarrhea, day care attendance, month of presentation and age at presen-Can J Infect Dis Vol 10 No 6 November/December 1999 425 Yield of stool testing of paediatric diarrhea   tation, only the latter two were associated with viral versus nonviral CAD (Table 5).In a separate logistic regression model for bacterial versus nonbacterial CAD vomiting, nausea, malaise, loss of appetite, watery versus bloody diarrhea, mucousy diarrhea, day care attendance, eating fast food recently, other family members with diarrhea, rash, abdominal tenderness, month at presentation and age at presentation were included.
As with the model for viral CAD, only the month and older age at presentation were associated with bacterial versus nonbacterial CAD (Table 5).Ten (66.7%) of the 15 laboratories that process stool specimens for paediatric hospitals across Canada (not including Alberta Children's Hospital) indicated that they still routinely receive requests for culture and sensitivity testing for bacteria other than C difficile in children with NAD.Eight (80%) of these 10 laboratories indicated that they still routinely perform such tests if requested, irrespective of duration of hospitalization.Five (38.5%) of 13 laboratories indicated that they have a policy, which was currently in practice, of only testing for C difficile toxin in such stool specimens.Two other hospitals were in the process of instituting this practice but did not have it in place at the time of the survey.Only one (6.67%) of the 15 laboratories routinely culture stool specimens for C difficile under any circumstances.Five (33.3%) of the 15 laboratories indicated that they still routinely receive requests for ova and parasite testing on NAD specimens.Three (60%) of these five routinely perform such tests if requested.

DISCUSSION
Recent studies, predominantly in adults, have questioned the value of routine screening of all stool specimens in nosocomial diarrhea (14,15,17,18,20,21).However, there has been very little information on this approach in paediatric nosocomial diarrhea.In previous studies conducted at our institution (20,21), we had noted that a single stool specimen was adequate for the identification of bacterial and parasitic pathogens in hospitalized children with diarrhea.Our current study indicates that testing for parasitic pathogens and bacteria other than C difficile is not necessary in children who have nosocomial diarrhea except in an outbreak setting.However, the telephone survey indicates that at least half of all paediatric centres surveyed still process stools from NAD patients for multiple pathogens.
The rate for infection associated NAD of 31.7/10,000discharges reported in our study is much higher than the average of 11.3/10,000 discharges reported in the National Nosocomial Infections Surveillance (NNIS) for the years 1985 to 1991.This may reflect actual higher rates in our hospital, or may reflect more complete laboratory evaluations of patients with diarrhea or differences in patient populations in reporting hospitals.It has previously been noted that NNIS rates are likely an underestimate because most participating hospitals do not have diagnostic virology laboratories (26).
The identified contribution of bacteria in this study of 19.4% for CAD and 23.3% for NAD is in the higher end of the documented rates of bacterial diarrheas in children of developed countries, which have ranged from 5.1% to 31% (6,(8)(9)(10)12,21,27,33,36,(42)(43)(44).Bacterial culture was positive in 10.6% CAD specimens, with E coli identified as the most common non-C difficile bacteria, present in 3% of CAD episodes.However, higher rates than ours have been reported for E coli (10% to 16.8%) (6,33,36,37,45), shigella (6% to 32.6%) (6,36,37,45,46), campylobacter (4.9% to 7.9%) (10,12), salmonella (4.1% to 11%) (8,9,10,18,34,36,37,46) and aeromonas (6.9%) (9).In only a few studies (37,45,46) was it clear that the population studies consisted only of CAD episodes.Two studies (45,46) took place in summer and fall, which biases their results in favour of bacterial isolation.The conclusions of the other studies may be explained if these episodes were cultured in the setting of an outbreak.The discrepancies could also be due to site to site differences because a study conducted by Church et al (21) at our institution revealed isolation rates for bacteria other than C difficile to be in the range of 0.2% to 2.3%, which are similar to the rates in our investigation.The identification of C difficile as the most common bacterial pathogen, although not unique (14,15,17,19), was notable in that it was the only bacterial organism identified in NAD, similar to the experience of Fan et al (18).
The association between C difficile diarrhea and antimicrobial exposure is well established (47)(48)(49)(50)(51).Although the role of C difficile as an etiological agent of diarrhea in the paediatric population is controversial, it cannot be excluded as a possible agent in children who have received previous antibiotic therapy and in whose stool other enteric pathogens have not been isolated (52).Previous authors have indicated that the incidence of community-acquired C difficile diarrhea is low, with less than one case/10,000 antibiotic prescriptions being reported from one large outpatient setting (48).Although we do not have similar comparisons with regards to rates per antibiotic usage, the high identification of C difficile in our NAD and CAD populations may reflect a high baseline rate of antibiotic usage, good detection techniques for C difficile or differences in our population in terms of food preparation and consumption.Similarly, patients with malignant disease receiving chemotherapy have been noted to be at risk for C difficile infection because of the numerous courses of antibiotics that they receive and because cancer chemotherapy itself may predispose patients to gut colonization with C difficile (53)(54)(55)(56).CAD episodes in which bacterial pathogens were isolated were predominantly in children older than five years of age.Caprioli et al (12) reported that stool specimens were positive for bacteria predominantly in children with a median age of 23 months, but failed to note how often stool specimens from older children were subjected to bacterial culture.
Our study confirms previous reports of low yield regarding bacterial stool culture (17,18) in children with NAD.A recent American study arrived at the same conclusions that bacterial cultures are not useful for nosocomial diarrhea (57).As in other published studies (14,15,17,19), positivity rates for C difficile toxin testing are higher than for other tests routinely performed on stool specimens.Screening for the presence of rotavirus antigen through EIA had relatively high yield for NAD episodes as well.Based on these findings, it would appear that testing for C difficile toxin and rotavirus antigen by the use of EIA are the only tests indicated in NAD episodes other than in the setting of an outbreak, where bacterial culture, and ova and parasite screening may be useful (58)(59)(60).Our study may be limited by the lack of universal use of EM at our institution, which may pick up Norwalk, toroviruses and other gastrointestinal viruses.These data need confirmation from other centres in Canada, in particular, the role of EM in routine management of nosocomial diarrhea needs clarification.
Ova and parasite examination in NAD may be indicated in the setting of compromised hosts where cryptosporidium may be implicated (61,62), or in inpatients who have multiple passes out of hospital and travel to local areas known to be endemic for parasites during these periods.There is controversy about the role of B hominis as a pathogen, and because it was identified only once along with rotavirus, its role as a cause of diarrhea in our study is dubious (63).
On univariate analysis, rotavirus was associated with day care attendance.This concurs with reported outbreaks of rotavirus-positive diarrhea in day cares (22,60).Vomiting, watery diarrhea and the absence of blood in the stool have been previously associated with rotavirus (4,8,23,35,41,46,64).In studies conducted in both adults (13) and children (12,44), blood in stool, abdominal pain and nausea were identified as good clinical predictors of a positive bacterial culture.Despite significant associations found on univariate analysis, our logistic regression model concluded that only the age of the patient and the month at presentation were associated with the isolation of either bacterial or viral agents.

CONCLUSIONS
There is a need to re-evaluate the performance of complete etiological examinations on stool specimens in nonoutbreak cases of NAD in children, a practice still common in Canadian paediatric centres.Screening for C difficile toxin and rotaviruses will identify the majority of infectious pathogens in NAD, while a broader range of organisms needs to be sought in CAD.The use of clinical-epidemiological factors such as age of the patient, seasonality, antibiotic use and travel history may be useful in helping to focus the testing performed on stool specimens.
age five years and older (76 of 219 episodes in children younger than age five years versus three of 40 episodes in children aged five years and older, P<0.001).Infants younger than 12 months of age formed the majority (55.9%) of the NAD patient population (P<0.001).Virus-positive specimens, and rotavirus specifically, were highest in the winter months (P<0.0001)(Figure1).Testing for C difficile toxin produced positive results in one-fourth of CAD cases and one-third of NAD episodes (Table 2).Bacterial pathogens other than C difficile were positive in 10.6% of CAD specimens sent, with E coli O157:H7 identified as the most common non-C difficile organism.In NAD episodes, there were no positive bacterial cultures (P<0.001,NAD versus CAD).In contrast with viruses, bacteria were most commonly identified in patients age five years and older(27 of 212 episodes in children younger than five years versus 31 of 74 episodes in children five years and older, P<0.0001) presenting with CAD.C difficile was the only bacterial pathogen identified in NAD episodes and was not prevalent in any one age group (P=0.38).C difficile was identified most often in 424 Can J Infect Dis Vol 10 No 6 November/December 1999 Deorari et al

TABLE 3 Clinical features of viral community-acquired diarrhea at the Alberta Children's Hospital, Calgary, Alberta from April 1, 1993 to March 30, 1995
*Viral episodes were those in which a viral pathogen was isolated from a stool; † Non-viral episodes were those in which viral pathogens were not isolated from a stool; episodes in which no tests performed on stool samples were excluded; ‡ Other family -other family members with diarrhea.MSK Musculoskeletal

TABLE 4 Clinical features of bacterial community-acquired diarrhea at the Alberta Children's Hospital, Calgary, Alberta from April 1, 1993 to March 30, 1995
*Bacterial episodes were those in which a bacterial pathogen was isolated from stool; † Nonbacterial episodes were those in which bacterial pathogen was not isolated from stool; excludes episodes in which no tests were performed on stool sample; ‡ Other family -other family members with diarrhea.MSK Musculoskeletal