Bauer TM, Lalvani A, Fehrenbach J, Steffen I, Aponte JJ, Segovia R, Vila J, Philippczik G, Steinbrückner B, Frei R, Bowler I, Kist M. Derivation and Validation of Guidelines for Stool Cultures for Enteropathogenic Bacteria Other Than Clostridium difficile in Hospitalized Adults. JAMA. 2001;285(3):313-319. doi:10.1001/jama.285.3.313
Author Affiliations: Department of Internal Medicine II (Drs Bauer, Fehrenbach, and Philippczik) and Institute of Medical Microbiology and Hygiene (Drs Steinbrückner and Kist), University Hospital, Freiburg, Germany; Nuffield Department of Clinical Medicine, University of Oxford (Dr Lalvani), and Department of Microbiology, John Radcliffe Hospital (Dr Bowler), Oxford, England; Bacteriological Laboratory, University Hospital Basel, Basel, Switzerland (Drs Steffen and Frei); and Departments of Microbiology (Dr Vila), Gastroenterology and Hepatology (Dr Segovia), and Epidemiology and Biostatistics Unit (Dr Aponte), Hospital Clínic, IDIBAPS, University of Barcelona, Barcelona, Spain.
Context The yield of in-hospital stool cultures performed more than 72 hours
after admission is low, and a commonly used policy dictates that laboratories
reject these cultures to save costs. However, enteropathogenic bacteria other
than Clostridium difficile (EPB) may cause nosocomial illness that would be missed by use of such
a "3-day rule."
Objective To develop guidelines for hospital use of stool cultures that are sensitive
to clinically relevant cases of sporadic and epidemic nosocomial diarrhea.
Design Five-part study that incorporated a derivation sample based on retrospective
chart review and a prospective cohort study (including cost savings analysis),
and a validation sample based on retrospective chart review.
Setting Four European academic health care centers.
Patients Derivation sample: 1735 adult inpatients from whom 3416 stool cultures
were obtained during a 19-month period (1995-1997) and 68 adult inpatients
for whom EPB were grown from stool cultures during a 10-year period (1988-1998);
validation sample: 65 patients with sporadic isolation of EPB (1993-1998),
56 patients involved in 2 nosocomial Salmonella outbreaks
(1992 and 1997), and 330 patients who had stool cultures performed (1998).
Main Outcome Measure Performance of derived criteria in detecting pathogenic bacteria and
outbreaks and reducing total number of stool cultures performed.
Results Stool cultures grew EPB in 3.3% of samples obtained ≤72 hours after
admission and 0.5% of samples obtained thereafter (P<.001).
Isolation of EPB >72 hours after admission was not associated with clinical
symptoms or signs but was associated with community-acquired diarrhea (24%),
age 65 years or older with preexisting comorbid disease (25%), neutropenia
(13%), HIV infection (10%), and nondiarrheal manifestations of enteric infections
(16%). Twelve percent were asymptomatic carriers. These characteristics were
used to create criteria for selecting patients for whom stool cultures would
be indicated. These criteria were applied post hoc to a series of 1025 stool
cultures; the number of stool cultures would have been reduced by 52% and
no clinically significant cases would have been missed. Annual savings to
a 355-bed institution would be approximately $7800 for reagent costs and 75
hours of technician time. In the validation samples, only 2 patients of 65
who had EPB would not have been identified, and neither required treatment.
If the 3-day rule had been applied, 52 cases would not have been identified,
28 of which required antibiotic treatment.
Conclusion Our modified 3-day rule for use in selecting cases for stool culture
is sensitive to sporadic and epidemic cases of nosocomial diarrhea in hospitalized
The yield of stool cultures performed in hospitalized patients is low.
Enteropathogenic bacteria other than Clostridium
species are grown from 2.6% to 6.4% of stool cultures1,2
and only 0.6% of stool cultures obtained more than 3 days after admission
(SC>3d).1,3 Cumulative laboratory
charges per positive result in hospitalized patients are approximately $1300
(1993 data),2 making a positive stool culture
report one of the most expensive microbiological results.4
Screening of stool samples for fecal leukocytes or blood may help reduce
the number of stool cultures processed in an outpatient setting,5
but is not feasible in hospitalized patients.2
Therefore, laboratory policies for rejecting stool cultures from hospitalized
patients have been implemented by a quarter of US microbiology laboratories.6 The most common policy is the so-called 3-day rule,
whereby SC>3d are generally rejected.2 Strict
implementation of this rule throughout the United States would result in savings
of $20 million to $73 million (data for 1990 and 1996, respectively).6,7 This policy is now receiving increasing
attention in Europe.3,8,9
Nosocomial diarrhea is most frequently an adverse effect of antibiotic
administration or enteral tube feeding or a consequence of Clostridium difficile infection.10 However,
"conventional" pathogens such as Salmonella, Yersinia enterocolitica, or Campylobacter may cause sporadic nosocomial diarrhea11
and nosocomial outbreaks.12,13
Prompt detection of such episodes would be compromised by the strict application
of the 3-day rule. Therefore, exceptions to this rule such as immunosuppression3 and suspected outbreaks14
have been proposed. Neither modification has been evaluated as to its practicability,
safety, and economic impact. Moreover, the definition of "immunosuppression"
varies widely, and nosocomial outbreaks may be difficult or even impossible
to suspect on clinical grounds.15,16
Therefore, we sought to develop a modified 3-day rule with precise criteria
for performance of stool cultures in hospitalized patients and to test whether
this modified 3-day rule would allow cost savings without negatively affecting
The University of Freiburg Hospital is an 1800-bed secondary and tertiary
care center and includes a 355-bed department of internal medicine with approximately
19 000 admissions and 121 000 inpatient hospital days per year.
The University Institute of Medical Microbiology and Hygiene performs approximately
2900 stool cultures in hospitalized patients per year, of which 74% are from
the department of internal medicine.
Stool samples are routinely plated onto Endo agar, Leifson agar, and
cefsulodin-irgasan-novobiocin agar. Additionally, 2 enrichment media for Salmonella species (Kauffmann broth and selenite broth)
are inoculated and subcultivated after 24 hours on Endo agar, Rambach agar,
and brillant green agar. Cefsulodin-irgasan-novobiocin agar plates are incubated
at 30°C for 48 hours, and all other media are incubated at 37°C for
24 hours. Samples submitted for the investigation of diarrhea are also inoculated
onto blood-free selective Campylobacter agar (incubated
at 37°C for 48 hours under microaerophilic conditions). Yersinia cold enrichment in saline (at 4°C for 7 days) is performed
when clinically indicated. Enterohemorrhagic and enteropathogenic Escherichia coli are searched for when clinically indicated by means
of a verotoxin enzyme-linked immunosorbent assay (Premier EHEC Test Kit, Meridian
Diagnostics, Cincinnati, Ohio) and serotyping, respectively. Organisms analyzed
for the purpose of this study were Salmonella species, Shigella species, Campylobacter
species, Yersinia species, enterohemorrhagic and
enteropathogenic E coli, and Vibrio species. Clostridium difficile–toxin
A/B is determined by an enzyme-immune assay (Ridascreen, R-Biopharm, Darmstadt,
The independent evaluation was performed in 3 European secondary and
tertiary academic health centers: University Hospital Basel, Basel, Switzerland
(900 beds, 1300 stool cultures from hospitalized patients per year), Hospital
Clínic, Barcelona, Barcelona, Spain (900 beds, 1200 stool cultures
per year), and the Oxford Radcliffe Hospitals NHS Trust, Oxford, England (1380
beds, 3700 stool cultures per year). These institutions had no standard culture
criteria in place at the time of the study and none routinely used empirical
treatment for nosocomial diarrhea. Patients with Salmonella or Shigella infection are generally treated
with antibiotics; patients with Campylobacter or Yersinia infections are treated on an individualized basis.
Because the study was observational with patient confidentiality maintained,
the University Hospital Freiburg institutional review board waived the requirement
for informed consent.
The study comprised 5 parts (Table
1). Part 1 was performed to provide baseline information about the
practice of ordering stool cultures and stool culture yield among adult (≥18
years) patients at the University of Freiburg Department of Internal Medicine.
All stool cultures and tests for C difficile toxin
(for comparison purposes) performed between November 1995 and October 1996
were analyzed retrospectively. Data collected included patient identification
number and date of admission, date of stool specimen, number of specimens
obtained during admission, and microbiological results.
Part 2 was a 7-month prospective survey in adult medical inpatients
for whom stool cultures had been ordered by ward staff, conducted to identify
patient characteristics that were predictors of stool culture positivity.
In addition to the data obtained in part 1, patient charts were reviewed on
the day their sample was received by the laboratory. Data evaluated included
a recent family history of diarrhea or recent travel outside of Europe, the
presence of nausea or vomiting, 8 or more bowel movements per day, temperature
exceeding 38°C, and presence of abdominal pain or tenderness.
Part 3 was a retrospective chart review designed to identify risk factors
for stool culture positivity in hospitalized adults. All patients from any
hospital department with a stool culture growing enteropathogenic bacteria
other than C difficile over a 10-year period (April
1988–October 1996 and June 1997–October 1998) were reviewed. Data
obtained were the same as those in part 1, plus the day of onset of diarrhea,
reason(s) for performing stool culture, preexisting comorbidity, use of immunosuppressive
drugs, and peripheral neutrophil count on the day the stool culture was performed.
In part 4, criteria for the performance of stool culture were defined
and applied post hoc to the data of part 2. Part 5 consisted of an independent
validation of the criteria at 3 different institutions. We assessed the sensitivity
of the criteria by performing 4- to 6 year reviews of all stool cultures growing
enteropathogenic bacteria other than C difficile
from hospitalized patients (analogous to part 3). We also studied the impact
of our criteria on the detection of nosocomial Salmonella outbreaks that had occurred in 2 study centers. We evaluated the efficiency
of our proposed criteria by retrospectively determining the percentage of
stool cultures that would not have been submitted under the premises of the
criteria during a consecutive 2-month period.
Comorbidity was defined as any preexisting disease that resulted in
permanently altered organ function, eg, cirrhosis, end-stage renal failure,
chronic obstructive pulmonary disease, active inflammatory bowel disease,
leukemia, or hemiparesis due to cerebrovascular accident.17
Immunosuppression was defined as human immunodeficiency virus (HIV) infection,
leukemia, malignant lymphoma, plasmocytoma, cirrhosis, diabetes mellitus,
end-stage renal failure, and use of cytotoxic or immunosuppressive drugs or
corticosteroids at dosages equivalent to or greater than 20 mg/d of prednisone.18 Nosocomial diarrhea was defined as the onset of 3
or more soft or liquid bowel movements at least 72 hours after admission,19 and neutropenia as a peripheral neutrophil count
of less than 0.5 × 109/L. The yield of stool culture was
defined as the proportion of stool cultures growing enteropathogenic bacteria
other than C difficile that had not been previously
reported in that patient, ie, first positive reports. The cost of reagents
needed for the processing of an average negative stool culture was calculated
on the basis of 100 consecutive negative stool cultures and amounted to $6.90
(EUR 4.62). Technician time required for the processing of an average negative
stool culture was estimated at 4 minutes.
Comparisons of proportions were done using the χ2 or
Fisher exact test where appropriate. Positivity rates from parts 1 and 2 were
combined using the Mantel-Haenszel method. Heterogeneity of positivity rates
between study periods was evaluated using the χ2 test for interactions.
During the 12-month and 7-month periods of parts 1 and 2, 2391 and 1025
stool cultures were analyzed, respectively, of which 21 (0.9%) and 13 (1.3%)
yielded first positive results of enteropathogenic bacteria other than C difficile. These positivity rates do not differ significantly
(P = .21), and therefore results of parts 1 and 2
were combined for the purpose of the following analyses. A total of 3416 stool
cultures from 1735 patients (mean age, 49 years; 49% men) during 1820 admissions
were included. First positive results were obtained from 34 stool cultures
(1.0%): Salmonella species in 17, Campylobacter species in 10, Yersinia species
in 6, and enterohemorrhagic E coli in 1 case. Eighty-five
percent of first positive results were obtained from the first specimen from
a given patient, the remainder from the second (5%) or third (10%) specimen.
A total of 2818 stool cultures (82.5%) were collected more than 72 hours
after admission (Figure 1). The
yield of SC>3d was more than 6-fold lower than for cultures taken within 72
hours of admission (20/598 [3.3%] vs 14/2818 [0.5%], P<.001).
Analysis of the clinical data obtained prospectively in part 2 failed to detect
any significant association between nausea or vomiting, 8 or more bowel movements
per day, temperature exceeding 38°C, and abdominal pain or tenderness
and the isolation of enteropathogenic bacteria other than C difficile. Subanalysis of stool cultures obtained within the first
3 days after admission revealed a significant association between a positive
result and a recent family history of diarrhea (P
= .02) or recent travel outside of Europe (P<.001).
During the 19-month period of parts 1 and 2, 2347 specimens from 1018
patients were submitted for C difficile–toxin
A/B tests (this test was not ordered for every patient). Of those, 89.6% were
obtained more than 72 hours after admission. Positivity rates for samples
obtained less and more than 72 hours after admission were 8.2% (20 cases)
and 5.5% (116 cases), respectively.
The 10-year review of part 3 identified 73 patients in whom SC>3d yielded
enteropathogenic bacteria other than C difficile.
Most cases occurred in the departments of internal medicine (68%) and surgery
(20%). Findings obtained in 68 patients for whom medical records could be
retrieved are shown in Table 2.
Nosocomial diarrhea due to enteropathogenic bacteria other than C difficile was diagnosed in 33 cases, the largest group being patients
aged 65 years or older with preexisting comorbidity. The following conditions
were found in this group: insulin-dependent diabetes mellitus with secondary
complications, cirrhosis, chronic renal failure, chronic obstructive pulmonary
disease, and quadriplegia. There were no positive SC>3d in any patient taking
immunosuppressive drugs whose peripheral neutrophil count was greater than
0.5 × 109/L.
We defined criteria for stool cultures in hospitalized adults that would
have identified all positive non–C difficile
SC>3d obtained during the 10-year period of part 3 (Table 3). These criteria and the traditional 3-day rule with and
without the exemption of immunosuppressed patients were retrospectively applied
to the patient series of part 2 to compare their impact on laboratory workload
and the ability to identify sporadic cases of nosocomial diarrhea due to enteropathogenic
bacteria other than C difficile (Table 4). Immunosuppression (as defined in the "Methods") was present
in 224 of 341 patients (65.7%) examined more than 72 hours after admission,
accounting for 74.4% of processed SC>3d. A total of 119 patients (34.9%) fulfilled
1 of our proposed criteria, accounting for 307 (36.7%) of SC>3d. Submission
of only those samples would have increased the yield of SC>3d from 0.8% to
2.0%. All first positive results would have been detected with the exception
of a 45-year-old patient who had undergone renal transplantation and had C difficile–associated diarrhea with an isolate of
uncertain pathogenicity (Y enterocolitica biovar
1 without virulence plasmid pYV).20
Reagent costs and technician time saved by omitting 63.3% of SC>3d in
the Freiburg 355-bed department of internal medicine would amount to approximately
$7800 (EUR 5200) and 75 hours annually, equivalent to total savings to the
hospital of approximately $10 500 (EUR 7000) annually. Based on US cost
estimates,6 the total reagent cost and technician
time savings would be approximately $10 300 and 356 hours annually.
Four- to six-year reviews of positive stool cultures from hospitalized
medical and surgical patients were performed at university hospitals in Basel,
Barcelona, and Oxford. Over a total cumulative period of 14 years during which
an estimated 27 000 stool cultures from hospitalized patients were collected,
65 patients were identified in whom enteropathogenic bacteria other than C difficile were grown from SC>3d (Table 5). Only 1 patient each from Basel and Barcelona was not covered
by the criteria: a 45-year-old woman following bone marrow transplantation
in whom Y enterocolitica was grown from a culture
obtained 4 days after recovery from neutropenia; and a 60-year-old man with
cirrhosis and melena with Campylobacter species grown
4 days after admission. Neither patient required treatment. Conversely, the
unmodified 3-day rule would have missed 52 patients with nosocomial diarrhea
related to enteropathogenic bacteria, 28 of whom required antibiotic treatment.
A series of 168 and 162 consecutive stool cultures obtained from hospitalized
patients over 2-month periods were reviewed in Basel and Barcelona, respectively
(Table 5). The percentage of SC>3d
collected that could have been avoided by application of the criteria was
47% and 62%, respectively.
Two food-borne nosocomial Salmonella outbreaks
were analyzed. The Basel outbreak of 1992 with S enteritidis affected 42 patients over 21 days, and the Barcelona outbreak of 1997
with Salmonella species of group C2 involved 14 patients
over 17 days. In both outbreaks the first 5 cases occurred on 5 nonadjacent
wards and therefore they were not suspected by ward staff. The outbreaks were
recognized by the microbiology laboratory after isolation of the second nosocomial Salmonella strain within a short period. Had the 3-day
rule been applied, the outbreaks would have been suspected at the earliest
after the second case on a given ward, which was the sixth case in both outbreaks.
This would have resulted in a delay of detection of at least 1 and 9 days,
corresponding to 6 and 34 excess patient-days of diarrhea in Basel and Barcelona,
respectively. Under the premises of our modified 3-day rule, 79% and 64% of
patients, respectively, would have had stool cultures even in the absence
of a suspected cluster. Therefore, both outbreaks would have been detected
after the third case. In Basel, this would not have delayed recognition because
cases 2 through 5 occurred on the same day. The Barcelona outbreak would have
been diagnosed 4 days later, causing an excess of 10 patient-days of diarrhea.
Our study demonstrates a low yield of stool cultures obtained from hospitalized
patients, confirming previous reports from the United States1,2,4,6,21
and various European countries.3,8,9,22
At the Freiburg University Hospital, enteropathogenic bacteria other than C difficile are grown from 0.5% of cultures obtained more
than 3 days after admission, compared with a 10-fold higher yield of tests
for C difficile–toxin A/B. Thus, testing for C difficile infection and consideration of noninfectious
causes such as antibiotic-associated23 or osmotic
diarrhea24 should be the first steps in the
evaluation of nosocomial diarrhea. However, enteropathogenic bacteria other
than C difficile may cause nosocomial diarrhea sporadically
in elderly or immunosuppressed patients or during hospital outbreaks.
We designed a comprehensive list of criteria for stool cultures that
would have identified all positive cultures in hospitalized patients over
a 10-year period. Retrospective application of the criteria to a consecutive
series of 1025 stool cultures showed that more than half of the cultures from
hospitalized patients could have been avoided, resulting in substantial cost
and time savings. Time savings were calculated based on laboratory time only
and do not reflect potential savings realized by clinicians and nurses. The
independent evaluation of this modified 3-day rule at 3 European university
hospitals confirmed its excellent safety profile and potential for cost savings.
Sporadic bacterial diarrhea may occur after nosocomial exposure to enteropathogenic
bacteria other than C difficile or following their
overgrowth during antibiotic treatment in previously asymptomatic carriers.25 We found 2 distinct forms of immunosuppression to
be associated with an increased risk for this occurrence: chemotherapy-associated
neutropenia and HIV infection. Exempting all other patients with immunosuppression
from the 3-day rule as suggested previously would greatly reduce if not eliminate
the economic impact of the culture policy without increasing its yield. A
third and hitherto unrecognized group of patients at risk is persons aged
65 years or older with preexisting comorbidity. The omission of SC>3d in all
patients aged 65 years or older would have impaired patient safety: 18 cases
of sporadic nosocomial bacterial enteritis in elderly patients at the Freiburg
University Hospital would have been missed over a period of 10 years. On the
other hand, no diagnosis of nosocomial diarrhea due to enteropathogenic bacteria
other than C difficile in patients aged 65 years
or older without preexisting comorbidity was identified in the retrospective
analyses performed in 4 centers, comprising approximately 50 000 stool
cultures from hospitalized patients.
Our review found that nosocomial diarrhea accounted for only half of
the positive SC>3d, and community-acquired diarrhea was the second largest
group. This may be due to delayed diagnostic workup and negative initial cultures:
up to 3 specimens may be necessary to detect 99% of pathogens.1,3
A third group of conditions yielding positive cultures at any time after admission
are nondiarrheal manifestations of enteric infection, such as mesenteric lymphadenitis,
acalculous cholecystitis,26 or extra-abdominal
signs such as reactive arthritis, erythema nodosum, or pyrexia of unknown
origin.27 Finally, stool cultures on selective
medium for Listeria species may be indicated when
a Listeria infection or a food-borne outbreak is
Outbreaks of nosocomial bacterial enteritis12
may cause high morbidity and mortality rates12
and require rapid recognition and treatment.28
They are readily recognized clinically when a large percentage of patients
is affected.28 However, episodes with low attack
rates may be identifiable only by the microbiology laboratory when an identical
pathogen is isolated from 2 or more seemingly sporadic cases of nosocomial
The detection of such outbreaks will inevitably be delayed by any kind of
laboratory rejection policy. The original 3-day rule would have significantly
delayed the recognition of the 2 outbreaks we analyzed. It is probable that
further nosocomial transmission and spread of the outbreak would have occurred
during this delay. In contrast, our modified 3-day rule addresses this problem
by routinely allowing for stool cultures in compromised patients who are likely
to develop symptomatic intestinal disease and require antibiotic treatment.33 The delay in the laboratory-based recognition of
these outbreaks would have been minimized and no vulnerable patients requiring
antibiotic treatment would have been missed.
In conclusion, we have defined criteria for stool cultures in hospitalized
patients that greatly reduce the laboratory burden of stool culture, yet provide
rapid diagnosis for patients at increased risk of nosocomial bacterial gastroenteritis.
Our criteria are based on readily available patient characteristics and can
be applied easily by ward staff ordering cultures. The applicability of these
criteria to other centers can be assessed by simple surveys. Inclusion of
these criteria into hospital guidelines and teaching programs may help refine
test ordering patterns by physicians, as has successfully been shown for previous
stool culture policies.14 Laboratory expenses
can thus be limited in an era of increasing cost constraint without compromising
individual patient care.