Fecal shedding of Salmonella spp in hospitalized horses can lead to outbreaks of nosocomial salmonellosis when adequate surveillance and infection control procedures are not used. The consequences of outbreaks of nosocomial Salmonella infections can be severe, resulting in potential infections in humans,1 fatalities in equine patients,1 loss of caseload and revenue,2 loss of teaching cases in university hospitals, and costs associated with cleaning, disinfection, and renovation of the hospital.2 Because the frequency of fecal shedding of Salmonella spp can be high in hospitalized horses, veterinary hospitals have instituted surveillance and infection control programs to reduce the risk of nosocomial infections.1–4
Identification of risk factors associated with nosocomial Salmonella infection in hospitalized horses is important so that effective control and preventative measures can be instituted to reduce the risk of disease transmission and potential outbreaks. Studies3,5,6 have provided an epidemiologic framework for investigation of risk factors associated with nosocomial Salmonella infections in hospitalized horses. In 2 studies5,6 conducted in the 1980s, horses treated with antimicrobial drugs, horses intubated with nasogastric tubes, and horses with an initial problem of colic at time of admission had an increased risk of nosocomial Salmonella infection. In a third study,3 diagnosis of large-colon impactions and duration of treatment with penicillin G potassium were associated with risk of nosocomial Salmonella infections. Although abdominal surgery and high caseload have been hypothesized as important contributors expected to increase the risk of nosocomial Salmonella infections, results of studies3,5,6 have failed to identify these 2 conditions as risk factors for nosocomial Salmonella infection in hospitalized horses. Research methods used in those studies were inconsistent and had several limitations. For example, in 2 studies,5,6 some control horses were tested for Salmonella organisms but others were not. In addition, in all 3 studies,3,5,6 it was not clear whether time of exposure to primary case horses was comparable between nosocomial case horses and control horses.
Currently, testing of horses for early detection of fecal shedding of Salmonella spp during hospitalization is a common practice in veterinary hospitals. This scenario provides an opportunity to reassess the epidemiologic aspects of nosocomial Salmonella infections in hospitalized horses. For example, it is possible that previous studies5,6 failed to identify abdominal surgery as a risk factor because many surgical inpatients were subclinically infected with Salmonella spp but were not detected because they were not tested; this subpopulation of horses could have been misclassified as susceptible control horses. The objectives of the study reported here were to examine the relationship between abdominal surgery and nosocomial Salmonella infections and the relationship between high caseload in combination with abdominal surgery and nosocomial Salmonella infections in horses hospitalized with signs of gastrointestinal tract disease.
Materials and Methods
Study population—All equine patients admitted to the University of Florida Veterinary Medical Center Large Animal Hospital between January 1, 2002, and December 31, 2006, with signs of gastrointestinal tract disease, as well as mares with foals ≤ 6 months old, were eligible for inclusion in the study. This subpopulation of horses was targeted for early detection of shedding of Salmonella organisms in feces as part of the hospital surveillance and infection control program. Horses that had non-fecal samples (such as blood, intestines, joint fluids, and abscesses) submitted for diagnosis of Salmonella spp were excluded from the study. Horses with incomplete data (missing information for Salmonella serogroup, serotype [unavailable or not typeable], antibiogram profile, admission date, discharge date, or sample collection date) were also excluded. In addition, horses hospitalized for < 72 hours or with < 2 fecal samples collected for bacterial culture were excluded.
Housing—At our large animal hospital facility, all horses admitted with signs of gastrointestinal tract disease (including those that undergo abdominal surgery) were housed in barn A. Horses that have a positive test result for Salmonella spp or were admitted with (or subsequently develop) diarrhea (defined as 3 consecutive bowel movements for which the feces do not sit on top of the bedding), fever (defined as rectal temperature > 39.2°C [102.6°F] when NSAIDs have not been administered or > 38.9°C [102.0°F] when NSAIDs have been administered during the preceding 24-hour period), or leukopenia (defined as WBC count < 5,000 cells/PL) were housed in an isolation unit. The intensive care unit was primarily used for patients that required critical care, including foals and their dams. Barn B was used for horses that were admitted because of lameness and ophthalmologicor reproductive-related problems or for elective procedures such as castration. Bovine and small ruminant patients were housed in a separate unit (ie, food animal barn) during hospitalization.
Hospital surveillance and infection control procedures—All equine patients with signs of gastrointestinal tract disease (as well as mares with foals ≤ 6 months old) admitted to our large animal hospital were targeted for early detection of shedding of Salmonella spp in feces at the time of admission and during hospitalization. A fecal sample was collected from the rectum of each horse within 12 hours after admission and submitted for bacterial culture; thereafter, additional samples were collected from the stall floor each morning prior to cleaning at 48-hour intervals until the patient was discharged from the hospital. Fecal samples collected outside regular business hours were refrigerated at 4°C prior to laboratory submission. For some horses, additional fecal samples were collected (eg, every 12 to 24 hours) at the discretion of the attending clinician. When a foal was admitted to the hospital with its dam, fecal samples were collected from both the foal and dam. A hospital infection control officer was responsible for overseeing the collection of fecal samples, microbiologic procedures, and collection and analysis of epidemiologic data. Any equine patient that had positive results for Salmonella spp or developed diarrhea, fever, or leukopenia was placed in the isolation unit. Isolation procedures included use of precautions (such as gloves, plastic boots, gowns, and footbaths) for personnel attending the patients.
Routine monthly environmental sampling for culture of Salmonalla spp was conducted at the hospital. Each month, samples were collected from 25 sites in various locations (isolation unit, intensive care unit, barn A, barn B, and food animal barn and the record rooms in the intensive care unit, barn A, and barn B); 1 or 2 sterile moist cellulose spongesa were used for sample collection at each site. In the isolation unit, 1 sponge was used to obtain a sample from the anteroom and a second sponge was used to obtain a sample from the stall. Within each site (eg, stalls), locations for sample collection included walls, doors, floors, and drains as well as cleaning tools and diagnostic equipment. In record rooms, locations for sample collection included the computer keyboard, computer mouse, telephone, and doorknobs. Samples were collected aseptically by the infection control officer, who wore sterile gloves (1 pair of gloves/sample). Each sponge was placed in a sterile plastic bag. The bag was sealed, labeled (sample identification number, date, and location), and submitted to our veterinary medical clinical microbiology laboratory. All 25 samples were collected and submitted to the laboratory on the same day. In addition to the monthly environmental samples, samples were collected from stalls used by horses that had positive results for Salmonella spp at admission or during hospitalization or horses with signs of gastrointestinal tract disease; these samples were collected after cleaning and disinfection procedures of the stalls were completed. These samples were labeled as quality-control samples and submitted for bacterial culture of Salmonella spp.
Laboratory and epidemiologic data were examined daily. Any time there was evidence of nosocomial infection or potential nosocomial infection, enhanced infection control measures were instituted immediately for all hospitalized patients. For example, use of footbaths, plastic boots, and gloves was mandatory for personnel attending each patient, and use of footbaths was instituted at all points of entry and exit in all barns and hallways within the hospital. The infection control officer established communication with all hospital personnel (clinicians, veterinary technicians, and veterinary medical students on clinical rotations) about the infection control status of the hospital. After cleaning and disinfection, environmental sampling procedures were conducted immediately to assess the magnitude of disease transmission and potential contamination of hospital facilities and equipment. Enhanced infection control measures were suspended after laboratory and epidemiologic data revealed no further evidence of nosocomial infection or hospital contamination.
Microbiologic procedures for detection of Salmonella organisms—Bacterial culture of fecal samples for detection of Salmonella organisms was performed at our veterinary medical clinical microbiology laboratory. For selective enrichment, 1 to 2 g of fresh feces was placed in 10 mL of selenite cystine broth,b and the solution was incubated at 37°C in an environment of 5% carbon dioxide for 24 hours. The 5% carbon dioxide environment was used because of limited incubator space; Salmonella spp are facultative anaerobes, and it was considered unlikely that 5% carbon dioxide would inhibit the growth of Salmonella organisms. The following day, a sample of the selenite cystine broth was subcultured on Hektoen enteric agar plates.c Plates were incubated at 37°C for 24 hours. Non–lactose fermenting, hydrogen sulfide–producing colonies were selected and isolated. These colonies were then inoculated on urea agar and lysine-iron agar slants and incubated at 37°C for 24 hours. Identification of urease-negative and hydrogen sulfide–producing organisms was established through use of a commercially available identification system.d Serogroup of Salmonella isolates was determined by means of agglutination; polyvalent (A through I and Vi) and group-specific (A through E) Salmonella O antiserae were used. Salmonella isolates were tested for antimicrobial susceptibility to amikacin, amoxicillin–potassium clavulanate, ampicillin-sulbactam, ampicillin, aztreonam, cefazolin, cefepime, cefotaxime, cefotetan, cefoxitin, ceftazidime, ceftizoxime, ceftriaxone, cefuroxime, cephalothin, chloramphenicol, ciprofloxacin, extended spectrum E-lactamase-a screen, extended spectrum E-lactamase-b screen, gatifloxacin, gentamicin, imipenem, levofloxacin, meropenem, nitrofurantoin, piperacillin, piperacillintazobactam, tetracycline, ticarcillin-clavulanate, tobramycin, and trimethoprim-sulfamethoxazole. Susceptibility patterns were determined by use of the minimum inhibitory concentration method7 with commercially prepared plates.f Interpretive breakpoints were based on guidelines published by the Clinical and Laboratory Standards Institute.7,8 Additional antimicrobials tested included cefpodoxime, ceftiofur, and enrofloxacin. The susceptibility patterns for these antimicrobials were determined by use of the Kirby-Bauer disk diffusion method.9 Serotyping of Salmonella isolates was performed at the USDA National Veterinary Services Laboratories in Ames, Iowa.
Primary cases—Horses that had positive results for Salmonella spp on fecal samples collected at the time of admission were classified as primary cases. In addition, horses with clinical signs of salmonellosis (ie, diarrhea, fever, or leukopenia) at the time of admission that had positive results for Salmonella spp on fecal samples collected during hospitalization (ie, on the second or subsequent samples) but with no evidence of nosocomial infection were also classified as primary cases. To rule out the possibility of nosocomial infection, surveillance data (Salmonella serotype, antibiogram profiles, stall location, and admission and discharge dates) as well as results of monthly environmental samples were reviewed to verify the classification of horses as primary cases.
Nosocomial cases—Hospital surveillance data (Salmonella serotype, antibiogram profile, admission and discharge dates, and housing location) were reviewed to classify horses as nosocomial cases. In this study, nosocomial cases were horses that had negative results for Salmonella spp in samples obtained at the time of admission and positive results ≥ 48 hours after admission. The source of nosocomial infection was a primary case horse that had positive results for Salmonella spp in samples obtained at the time of admission or during hospitalization and that shared the same serotype and antibiogram as the nosocomial case horse. In addition, there was an overlap between admission and discharge of the primary case and nosocomial case horses. Another source of nosocomial infection was environmental contamination. Horses that had negative results for Salmonella spp in samples obtained at the time of admission but had positive results thereafter and that shared the same Salmonella serotype and antibiogram profile as a Salmonella-positive environmental sample collected during the period of hospitalization were also classified as nosocomial cases. When environmental contamination was attributed as the source of infection, the nosocomial case horse was never exposed to the primary case horse associated with the environmental contamination (ie, there was no overlap from admission to discharge between the primary case and nosocomial case horses because the primary case horse had already been discharged from the hospital or euthanized before the nosocomial case horse was admitted).
Susceptible (control) horses—Horses that had negative results for Salmonella spp on fecal samples collected within 12 hours after admission and on all subsequent samples collected every 48 hours during hospitalization were classified as control horses.
Study design—The study was designed as a matched case-control study. To accomplish objective 1 (examine the relationship between abdominal surgery and nosocomial Salmonella infections), 1 to 4 control horses were matched to each nosocomial case horse on the basis of admission date (± 2 days) of the primary case or on the basis of the collection date (± 2 days) of the Salmonella-positive environmental sample. We considered matching on the basis of admission date (± 1 day) of the primary case or the collection date (± 1 day) of the Salmonella-positive environmental sample, but the number of eligible control horses did not enable us to achieve a 1:1 ratio of control horses to case horses. In addition, we considered matching on the basis of admission date of the nosocomial case horses, but examination of medical records revealed that many control horses were never exposed to the primary case horse (eg, hospital duration was not the same in nosocomial case and control horses). A total of 28 horses were classified as nosocomial cases; environmental contamination was considered the source of infection for 5 case horses. Twelve of the 28 nosocomial case horses were excluded (5 because of the lack of matching control horses and 7 because of missing information in the medical records); 4 of these 12 case horses underwent abdominal surgery during hospitalization. Final enrollment for objective 1 included 16 nosocomial case and 35 control horses. The frequency of abdominal surgery and other investigated exposure factors was compared between case and control horses.
To accomplish objective 2 (examine the relationship between high caseload in combination with abdominal surgery and nosocomial Salmonella infections), 4 control horses were matched to each nosocomial case horse on the basis of year of admission. Seven of the 28 nosocomial case horses were excluded for objective 2 because of missing information in the medical records. Final enrollment for objective 2 included 21 case and 84 control horses. A computer programg was used to generate random numbers (ie, select horses) needed in this portion of the study.
Data collection—A questionnaire was developed for collection of epidemiologic data, including patient age, sex, breed, admission date, season of the year at admission (summer was defined as May through September, and winter was defined as October through April), discharge date, duration of hospitalization, number of fecal samples collected for bacterial culture of Salmonella spp during hospitalization, caseload (total number of patients at our large animal hospital, including equids, cattle, goats, pigs, and camelids) at admission as well as during the 5-day period that included 4 days before and the day of admission, number of horses shedding Salmonella spp in feces at admission and during the 5-day period that included the 4 days before and the day of admission, number of days of hospitalization during which both the nosocomial case and control horses were exposed to the primary case horse (eg, duration of overlap between admission and discharge dates), clinical findings during hospitalization (including reason for hospitalization, surgical procedures, clinical procedures [nasogastric intubation, rectal palpation, and abdominocentesis]), use of anti-inflammatory agents, and use of sedatives, analgesics, or antimicrobial drugs during hospitalization. In addition, the number of patient–care personnel contacts during the first 3 days after admission were recorded by examination of medical records; a 3-day window for time of exposure was selected on the basis of the median distribution for the number of days a horse was hospitalized until the first sample was collected that yielded a positive result. Number of patient–care personnel contacts was determined by counting the number of times hospital (care) personnel entered a patient's stall to perform procedures such as administration of medication, IV administration of fluids, gastric decompression, and measurement of vital signs (ie, rectal temperature and heart rate). Number of contacts was determined from the time a patient was placed into a stall (where it was housed during hospitalization) until the time the patient was discharged. At our large animal hospital, every time hospital personnel entered a patient's stall, the time and reason for entering the stall were recorded. Contact prior to placement in a stall (including physical examination during initial assessment at admission and during surgery) was not measured. Finally, for nosocomial case horses, the number of days hospitalized until the first sample was collected that yielded positive results and the serotype of the Salmonella spp were recorded.
Statistical analysis—Conditional logistic regression was used to model the odds of being a nosocomial case horse as a function of investigated risk factors.10,h Initial screening of potential risk factors for nosocomial Salmonella infection was performed by use of univariable conditional logistic regression or the Wilcoxon rank sum test11,i for continuous variables such as caseload, number of horses shedding Salmonella spp in feces, time of exposure (number of days) to the primary case horse, number of fecal samples collected from each horse, duration of hospitalization, and number of patient–care personnel contacts. Initially, variables associated (P ≤ 0.20) with the outcome of interest (nosocomial infection) were entered into the model, and a forward stepwise approach was used to identify variables associated with infection by use of 2-sided P-values-to-enter and P-values-to-remove of 0.05 and 0.10, respectively. Duration of hospitalization and number of fecal samples collected from each horse for bacteriologic culture of Salmonella spp were included as required variables in the final model (objective 1) because they could have influenced the probability of detecting Salmonella organisms. The interaction terms of abdominal surgery X use of antimicrobial drugs and abdominal surgery X patient–care personnel contacts were examined in the final model. Values for the final model were considered significant at P ≤ 0.05. Similarly for objective 2, the variables for number of horses shedding Salmonella spp at admission and number of fecal samples collected from each horse were included as required variables in the final model. The interaction term for caseload X abdominal surgery was included in the final model and tested for significance. Fit of the model to the data was assessed by a visual examination of residual plots (standardized Δ-β values vs observation number and Δ-β vs fitted values). Case-control sets that had horses with extreme Δ-β values and low fitted values were excluded from the analysis to evaluate their influence on estimated ORs. In the final model, the adjusted OR and 95% CI were reported. The OR was used as an epidemiologic measure of association between a risk factor and risk of nosocomial Salmonella infection. Thus, when a particular factor was not associated with risk of infection, the OR was 1. The greater the departure of the OR from 1 (ie, values larger or smaller than 1), the stronger the association was between the factor (eg, abdominal surgery) and risk of nosocomial Salmonella infection.
For objective 1, the null hypothesis that number of patient–care personnel contacts during the first 3 days after admission was not different between horses that underwent abdominal surgery and horses that did not was tested by use of the median test. For objective 2, the null hypothesis that number of horses shedding Salmonella spp at admission was not different during times of low (15 to 25 hospitalized patients) or high (26 to 54 hospitalized patients) caseload was tested by use of the median test.11 The variable for caseload was classified into 2 groups by use of the median distribution to be consistent with the multivariable analysis and to simplify the interpretation of the OR.
Results
Objective 1—Sixteen horses were classified as nosocomial case horses and 35 as control horses. Of the 16 case horses, 3 had diarrhea, fever, and leukopenia; 1 had diarrhea and fever; 2 had diarrhea only; 2 had fever and leukopenia; 1 had leukopenia only; 5 did not have diarrhea, fever, or leukopenia; and 2 did not have diarrhea or fever but results for leukopenia were not known. Salmonella enterica serotype Newport (n = 5) was the most commonly isolated serotype among nosocomial cases, followed by serotypes Reading (3), Anatum (2), Branderup (1), Java (1), Javiana (1), Litchfield (1), Meleagridis (1), and Saint-paul (1). There were 5 isolates in Salmonella serogroup B, 1 in serogroup C1, 6 in serogroup C2, 1 in serogroup D, and 3 in serogroup E.
The fecal sample that yielded the first positive result for Salmonella spp was the second (n = 5 horses), third (7), fourth (2), fifth (1), or sixth (1) sample collected. The fecal sample that yielded the first positive result for Salmonella spp was collected on Monday (n = 5), Tuesday (1), Wednesday (3), Thursday (1), or Friday (6). The median number of days a nosocomial case horse was hospitalized until the first fecal sample was collected that yielded positive results was 3 days (first and third quartiles, 2 and 4 days, respectively).
Eleven of the 16 Salmonella serotypes isolated from nosocomial cases were susceptible to all antimicrobial drugs. Of the 5 Salmonella isolates that had antimicrobial resistance, 2 were from horses hospitalized in early to mid January 2002 and 3 were from horses hospitalized in mid November to early December 2004. The Salmonella spp isolated in 2002 were serotype Java (resistant to amoxicillin–potassium clavulanate, ampicillin, cefazolin, ceftiofur, clindamycin, erythromycin, oxacillin, penicillin, rifampin, doxycycline, tetracycline, and trimethoprim-sulfamethoxazole) and serotype Newport (resistant to clindamycin, erythromycin, oxacillin, penicillin, and rifampin). The 3 antimicrobial-resistant Salmonella spp isolated in 2004 were all serotype Reading, and all 3 were resistant to amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, and tetracycline.
Median number of days exposed to the primary case horse was 4 days for nosocomial case horses and 3 days for control horses; these values did not differ significantly (P = 0.13; Table 1). Median number of fecal samples collected did not differ significantly (P = 0.07) between nosocomial case horses (4 samples) and control horses (3 samples). Median number of days in the hospital did not differ significantly (P = 0.16) between nosocomial case horses (7 days) and control horses (5 days). Median number of patient–care personnel contacts during the first 3 days after admission did not differ significantly (P = 0.61) between nosocomial case horses (29 contacts) and control horses (28 contacts).
Caseload, number of horses shedding Salmonella spp at admission, number of fecal samples collected, and duration of hospitalization in nosocomial case and control horses (objective 1).
Variable | Case horses (n = 16) | Control horses (n = 35) | P value* |
---|---|---|---|
Caseload (No. of hospitalized animals) | |||
At admission | 24(19,28) | 29 (23,30) | 0.19 |
At admission and the 4 days before admission | 124(107,140) | 129(113,143) | 0.85 |
No. of horses shedding Salmonella spp | |||
At admission | 3 (2,4) | 4 (3,5) | 0.15 |
At admission and the 4 days before admission | 4(3,6) | 4(3,7) | 0.38 |
Mean No. of horses shedding Salmonella spp at admission and the 4 days before admission | 0.7(0.6,1.4) | 0.8(0.6,1.4) | 0.42 |
Duration of exposure to primary case horse (d) | 4(2,7) | 3(2,4) | 0.13 |
No. of fecal samples collected | 4(3,6) | 3(2,4) | 0.07 |
Duration of hospitalization (d) | 7(4,11) | 5(4,8) | 0.16 |
No. of patient-care personnel contacts during the first 3 days after admission | 29 (24,39) | 28(18,40) | 0.61 |
Data reported are median (first, third quartiles).
Values were considered significant at P ≤ 0.05.
Eight of the 16 nosocomial case horses underwent abdominal surgery. Of the 8 horses that underwent abdominal surgery, 6 did not develop diarrhea during hospitalization.
Use of univariable conditional logistic regression analysis revealed that patient age, withholding of feed, abdominal surgery, use of sedatives, and use of antimicrobial drugs had values of P ≤ 0.20 and were included in the multivariable analysis (Tables 2 and 3). The variables duration of hospitalization, number of fecal samples collected, and number of patient–care personnel contacts during the first 3 days after admission were forced into the final model.
Frequency distribution of host factors (age and sex), history of antimicrobial use, initial primary concern, housing during hospitalization, crude OR, and 95% Cl of investigated risk factors among case and control horses (objective 1).
Variable | Case horses (n = 16) | Control horses (n = 35) | OR | 95% Cl | P value* |
---|---|---|---|---|---|
Age | |||||
Foal ≤ 6 months old | 5 | 4 | 1.00 | Reference | NA |
Adult | 11 | 31 | 0.24 | 0.04, 1.36 | 0.10 |
Sex | |||||
Female | 6 | 16 | 1.00 | Reference | NA |
Male | 7 | 6 | 2.74 | 0.67, 11.13 | 0.15 |
Gelding | 3 | 13 | 0.50 | 0.08, 2.95 | 0.45 |
History of antimicrobial drug use prior to admission | |||||
No | 14 | 23 | 1.00 | Reference | NA |
Yes | 2 | 6 | 0.62 | 0.11, 3.45 | 0.59 |
Initial primary concern | |||||
Colic | 12 | 21 | 1.00 | Reference | NA |
Fever | 1 | 2 | 0.78 | 0.06, 8.87 | 0.84 |
Diarrhea | 2 | 0 | ND | ND | ND |
Anorexia | 0 | 1 | ND | ND | ND |
Pneumonia | 0 | 2 | ND | ND | ND |
Other† | 1 | 8 | 0.20 | 0.02,1.87 | 0.15 |
Housing | |||||
Isolation barn | 3 | 2 | 1.00 | Reference | NA |
Intensive care unit | 2 | 8 | 0.32 | 0.03, 3.23 | 0.34 |
Barn A | 11 | 24 | 0.49 | 0.09, 2.63 | 0.40 |
Barn B | 0 | 1 | ND | ND | ND |
Duration of hospitalization (d) | |||||
3 to 5 | 5 | 22 | 1.00 | Reference | NA |
≥ 6 | 11 | 13 | 4.43 | 0.88, 22.36 | 0.07 |
Reference was the category against which other categories were compared in the analysis.
Other includes corneal ulcer, uveitis, umblical hernia, a clinically normal mare with a sick foal, and a sick mare with a clinically normal foal.
Frequency distribution of hospital procedures, clinical findings, diagnosis, crude OR, and 95% Cl of investigated risk factors among case and control horses (objective 1).
Variable | Case horses (n = 16) | Control horses (n = 35) | OR | 95% Cl | P value* |
---|---|---|---|---|---|
Nasogastric intubation findings | |||||
Normal | 2 | 8 | 1 | Reference | NA |
Abnormal | 6 | 9 | 1.88 | 0.31, 11.41 | 0.48 |
Not done | 8 | 18 | 1.57 | 0.28, 8.65 | 0.59 |
Rectal examination findings | |||||
Normal | 2 | 7 | 1 | Reference | NA |
Abnormal | 7 | 21 | 0.8 | 0.08, 7.25 | 0.84 |
Not done | 7 | 7 | 3.78 | 0.42, 33.94 | 0.23 |
Abdominocentesis findings | |||||
Normal | 1 | 8 | 1 | Reference | NA |
Abnormal | 4 | 3 | 9.61 | 0.5, 182.79 | 0.13 |
Not done | 11 | 24 | 3.73 | 0.34, 41.09 | 0.28 |
Withhold feed before surgery | |||||
No | 4 | 17 | 1 | Reference | NA |
Yes | 12 | 18 | 2.54 | 0.67, 9.5 | 0.16 |
Abdominal surgery | |||||
No | 8 | 29 | 1 | Reference | NA |
Yes | 8 | 6 | 4.29 | 1.07, 17.12 | 0.03 |
Anti-inflammatory use | |||||
No | 3 | 10 | 1 | Reference | NA |
Yes | 13 | 25 | 1.19 | 0.22, 6.44 | 0.83 |
Sedative or analgesic use | |||||
No | 3 | 17 | 1 | Reference | NA |
Yes | 12 | 18 | 4.5 | 0.86, 23.43 | 0.07 |
Not known | 1 | 0 | ND | ND | ND |
Antimicrobial use | |||||
No | 4 | 22 | 1 | Reference | NA |
Yes | 12 | 13 | 5.92 | 1.22, 28.64 | 0.02 |
Diarrhea† | |||||
No | 10 | 26 | 1 | Reference | NA |
Yes | 6 | 9 | 1.6 | 0.43, 5.87 | 0.47 |
Fever‡ | |||||
No | 10 | 26 | 1 | Reference | NA |
Yes | 6 | 9 | 1.79 | 0.47, 6.74 | 0.38 |
Leukopenia§ | |||||
No | 8 | 21 | 1 | Reference | NA |
Yes | 6 | 4 | 3.92 | 0.70, 21.97 | 0.11 |
Not known | 2 | 10 | 0.55 | 0.09, 3.22 | 0.5 |
No. of patient-care personnel contacts during first 3 days | |||||
7 to 28 | 7 | 19 | 1 | Reference | NA |
≥ 29 | 9 | 16 | 1.24 | 0.31, 4.82 | 0.75 |
Diagnosis | |||||
Small-colon impaction | 0 | 3 | ND | ND | ND |
Large color impaction | 0 | 5 | ND | ND | ND |
Entoritic | 5 | 0 | ND | ND | ND |
Deal impaction | 0 | 5 | ND | ND | ND |
Large-color displacement | 3 | 0 | ND | ND | ND |
Other ∥ | 8 | 21 | ND | ND | ND |
Diarrhea was defined as 3 consecutive bowel movements for which the feces do not sit on top of the bedding.
Fever was defined as rectal temperature >39.2°C (102.6°F) when NSAIDs have not been administered or >38.9°C (102.0°F) when NSAIDs have been administered during the preceding 24-hour period.
Leukopenia was defined as WBC count < 5,000 cells/μL.
Other includes ileal strangulation, rotavirus-induced diarrhea, Rhodococcus-induced pneumonia, cecal impaction, corneal ulcer, nephrosplenic entrapment, ileocecal intussusception, enterolithiasis, gastric ulcer, umbilical hernia, colon torsion, peritonitis, a clinically normal mare with a sick foal, and a sick foal with a clinically normal mare.
See Tables 1 and 2 for remainder of key.
In the multivariable analysis, addition of the 2-way interaction terms abdominal surgery X use of antimicrobial drugs and abdominal surgery X patient–care personnel contacts did not contribute to the final model; thus, these terms were removed from the model. After controlling for duration of hospitalization, number of fecal samples collected, and number of patient–care personnel contacts during the first 3 days after admission, horses that underwent abdominal surgery were significantly (P = 0.03) more likely to become infected with Salmonella spp during hospitalization than horses that did not undergo surgery (OR = 8.2; 95% CI = 1.11, 60.24; model 1; (Table 4). Visual examination of residuals revealed that the Δ-β values for the variable abdominal surgery were not extreme (ie, not > 1), which supported overall goodness-of-fit. Analysis of residuals (set of case and control horses with the largest Δ-β value and the lowest fitted value) indicated the existence of influential observations; however, removal of these observations did not change the finding of greater risk associated with abdominal surgery.
Results for multivariable conditional logistic regression models for nosocomial Salmonella infections in hospitalized horses (objective 1).
Analysis | Variable | Adjusted OR | 95% Cl | P value* |
---|---|---|---|---|
Model 1 | Abdominal surgery | |||
No | 1.00 | Reference | NA | |
Yes | 8.20 | 1.11, 60.24 | 0.03 | |
Duration of hospitalization (d) | ||||
3 to 5 | 1.00 | Reference | NA | |
≥6 | 1.05 | 0.10, 10.73 | 0.96 | |
No. of fecal samples collected | 1.57 | 0.82, 3.00 | 0.17 | |
No. of patient-care personnel contacts during first 3 days after admission | 0.99 | 0.94,1.03 | 0.68 | |
Model 2 | No abdominal surgery and no antimicrobial drug use | 1.00 | Reference | NA |
No abdominal surgery but antimicrobial drug use | 5.04 | 0.28, 89.11 | 0.26 | |
Abdominal surgery and antimicrobial drug use | 15.35 | 1.35, 173.93 | 0.02 | |
Duration of hospitalization (d) | ||||
3 to 5 | 1.00 | Reference | NA | |
≥6 | 1.06 | 0.11, 9.59 | 0.95 | |
No. of fecal samples collected | 1.35 | 0.69, 2.65 | 0.37 | |
No. of patient-care personnel contacts during first 3 days after admission | 0.97 | 0.91, 1.02 | 0.32 |
See Tables 1 and 2 for key.
Because all horses that underwent abdominal surgery were treated with antimicrobial drugs, a second model was generated to examine the risk of infection in this group of horses, the group of horses that did not have surgery but were treated with antimicrobial drugs, and the horses that did not have either exposure (model 2; Table 4). In model 2, the odds of infection were significantly (P = 0.02) higher (15 times as high) in horses that underwent abdominal surgery and were treated with antimicrobial drugs, compared with the odds for horses that did not have either of these factors (OR = 15.35; 95% CI = 1.35, 173.93). The odds of infection were 5 times as high in horses that did not have abdominal surgery but were treated with antimicrobial drugs, but this association (OR = 5.04; 95% CI = 0.28, 89.11) was not significant (P = 0.26).
Median number of patient–care personnel contacts during the first 3 days after admission was 32 in horses that underwent abdominal surgery and 26 in horses that did not; these values differed significantly (P = 0.04). Conditional logistic regression was used to examine the potential interaction effect between abdominal surgery and number of patient–care personnel contacts, but this effect was not relevant (OR = 0.96; 95% CI = 0.78, 1.17) or significant (P = 0.71). Finally, 4 of the 12 nosocomial case horses that were excluded from the analysis because of a lack of matching control horses or missing information in the medical records underwent abdominal surgery. Had these 12 horses been included in the analysis, the prevalence of abdominal surgery among case horses would have decreased from 50% (8/16) to 42% (12/28), compared with 17% (6/35) among control horses. We conducted another analysis in which we used all 28 case and 35 control horses, and the odds of infection associated with abdominal surgery decreased from 4.2 to 3.6 (decrease of 16%), but the association (OR = 3.6; 95% CI = 1.4, 11.5) remained significant (P = 0.02).
Objective 2—Twenty-one horses were classified as nosocomial case horses and 84 as control horses. Caseload (P = 0.91) or season of the year at admission (P = 0.99) were not significantly associated with risk of nosocomial Salmonella infection. After controlling for number of horses shedding at admission and number of samples collected, high caseload in combination with abdominal surgery was not identified as a significant (P = 0.35) risk factor for nosocomial Salmonella infection in hospitalized horses (OR = 2.81; 95% CI = 0.31, 25.32), as determined by use of a third model (model 3; Table 5). Abdominal surgery was confirmed as a significant (P = 0.01) risk factor associated with nosocomial Salmonella infection (OR = 4.98; 95% CI = 1.55, 15.95), as determined by use of another model (model 4).
Results for multivariable conditional logistic regression models for nosocomial Salmonella infections in hospitalized horses (objective 2).
Analysis | Variable | Adjusted OR | 95% Cl | P value* |
---|---|---|---|---|
Model 3 | Hospital caseload | |||
Low (15 to 25 patients) | 1 | Reference | NA | |
High (26 to 54 patients) | 0.56 | 0.13, 24.32 | 0.42 | |
Abdominal surgery | ||||
No | 1 | Reference | NA | |
Yes | 2.95 | 0.60, 14.41 | 0.18 | |
Hospital caseload × abdominal surgery interaction | 2.81 | 0.31, 25.32 | 0.35 | |
Model 4 | Hospital caseload | |||
Low (15 to 25 patients) | 1 | Reference | NA | |
High (26 to 54 patients) | 0.84 | 0.26, 2.67 | 0.77 | |
Abdominal surgery | ||||
No | 1 | Reference | NA | |
Yes | 4.98 | 1.55,15.95 | < 0.01 |
The analyses were adjusted for number of horses shedding Salmonella spp at admission and number of fecal samples collected.
See Tables 1 and 2 for remainder of key.
Median number of horses shedding Salmonella spp in feces at admission was not significantly (P = 0.65) different during periods of high caseload (26 to 54 hospitalized patients; 2 shedders) or low caseload (15 to 25 hospitalized patients; 2 shedders).
Environmental samples—Environmental contamination was considered the source of infection in 5 of 28 horses initially classified as nosocomial case horses. Salmonella Reading was recovered initially from a fecal sample of one of the aforementioned nosocomial case horses and subsequently from a sample obtained from the stall used by that horse. The environmental sample was a quality-control sample obtained 2 days after the nosocomial case horse was discharged; it was obtained to assess the efficacy of the cleaning and disinfection procedures. Interestingly, the isolate from the environmental sample from the stall had a different pattern of antimicrobial resistance (resistant to ceftiofur in addition to amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, and tetracycline) from that of the isolate from the fecal sample from the horse. Another horse was admitted and placed in a different stall but in the same barn 2 days after the nosocomial case horse was discharged, and that horse had a positive result for Salmonella Reading 2 days after it was placed in the stall (4 days after discharge of the nosocomial case horse). However, this horse was excluded from the case-control study (objective 1) because of a lack of a matching control horse.
Discussion
The study reported here provided epidemiologic evidence that equine patients hospitalized with gastrointestinal tract disease and that undergo abdominal surgery are at high risk of nosocomial Salmonella infection, compared to horses hospitalized with gastrointestinal tract disease that do not undergo abdominal surgery. However, the study results do not support the hypothesis that hospital caseload alone or in combination with abdominal surgery is a predisposing factor for nosocomial Salmonella infection in hospitalized horses at our large animal hospital.
The strengths of the study are the inclusion of nosocomial case and susceptible (control) horses exposed to primary case horses, the use of control horses defined as Salmonella-culture-negative horses, the number of fecal samples collected and tested, and the duration of hospitalization (which was comparable between nosocomial case and control horses). The small number of nosocomial case horses is a study limitation that affected the precision of the risk estimates of nosocomial infections associated with abdominal surgery. The small sample size also affected our ability to adequately assess the potential interaction effects of caseload and abdominal surgery on risk of nosocomial Salmonella infections. Finally, the study population was restricted to equine patients admitted with primary clinical signs of gastrointestinal tract disease and mares with foals ≤ 6 months old. Thus, our study results cannot be generalized to the entire equine population at the University of Florida large animal hospital.
In this study, 1 to 2 g of fresh feces was placed in 10 mL of selenite cystine broth (ratio, 1:10 to 1:5) and the broth was incubated at 37°C overnight for selective enrichment of Salmonella spp. This laboratory procedure differs from that reported in another study3 in which investigators placed 5 g of fresh feces in 100 mL of selenite broth (ratio, 1:20) and incubated the broth at 37°C overnight. Laboratory procedures for culture of Salmonella spp are not clearly standardized among veterinary microbiology laboratories.12 This reality can further affect the comparison of study results reported here and those in other epidemiologic studies3,5,6 in which investigators evaluated risk factors associated with nosocomial Salmonella infections. In 1 study,13 weight of the fecal sample obtained from pigs had an effect on detection of Salmonella spp; sensitivity increased from 32% to 63% when weight of the fecal sample increased from 1 to 10 g (diluted 1:9 [wt:wt] with buffer peptone water solution). It is difficult to know how those results in swine can be extrapolated to the horses in our study. The study in swine used only 1 sample/pig. In our study, the median number of samples collected was 3 for control horses and 4 for nosocomial case horses. Assuming the concentration of Salmonella spp in the first fecal sample collected in our study population was similar to that in swine,13 it is possible our surveillance system may have missed Salmonella shedders at admission because of low sensitivity. However, risk of misclassification of primary cases (false-negative results) in our study was reduced because horses with clinical signs of salmonellosis at admission (ie, diarrhea, fever, or leukopenia) that had positive results for Salmonella spp on fecal samples collected later during hospitalization (ie, the second, third, or subsequent samples) but with no evidence of nosocomial infection were classified as primary cases. In addition, risk of misclassification of nosocomial cases (false-negative results) was reduced in the study because the median number of fecal samples collected for control horses was 3 (rather than 1). Finally, horses can shed Salmonella organisms intermittently,14 and it is currently recommended that at least 5 fecal samples be submitted for microbiologic culture to have > 95% confidence that a horse is negative for Salmonella spp.14–16 In our study, the median number of fecal samples collected and submitted for microbiologic culture for horses classified as control horses was 3. This number is fewer than the recommended 5 fecal samples required to classify a horse as negative for Salmonella spp. Therefore, it is possible that some infected horses may have been misclassified as control horses. Assuming that some nosocomial case horses were misclassified as control horses, and the frequency of the main exposure of interest (abdominal surgery) in this group of horses was similar to that for the nosocomial case horses (ie, 50%), then the risk of nosocomial infection as a result of abdominal surgery may have been underestimated.
In the study reported here, horses that underwent abdominal surgery were more likely to become infected with Salmonella spp during hospitalization. To our knowledge, this is the first observational study that provides evidence that abdominal surgery is a risk factor for nosocomial Salmonella infection in hospitalized horses. Three studies3,5,6 failed to identify abdominal surgery as a risk factor for nosocomial Salmonella infection. In 2 of those studies,5,6 samples were not routinely collected from control horses and submitted for culture to identify Salmonella spp; therefore, some horses classified as control horses may have been primary or nosocomial case horses. In our study, 8 of 16 nosocomial case horses underwent surgery, and 6 of those 8 horses did not develop diarrhea during hospitalization. In the aforementioned 2 studies,5,6 this subpopulation of horses could have been classified as control horses, which makes it difficult to identify abdominal surgery as a risk factor. Finally, in the third study,3 similar to the other 2 studies,5,6 it was not clear whether duration of exposure to primary case horses was comparable between nosocomial case and susceptible (control) horses. Thus, it is difficult to interpret the epidemiologic analysis of that study,3 especially for abdominal surgery as a risk factor.
Horses that undergo abdominal surgery have substantial amounts of stress.16 The surgical procedure causes stress to patients; in addition, the large colon or cecum may be evacuated and lavaged, feed may be withheld or changed, antimicrobial drugs may be administered, and various degrees of ileus may develop.3,14,17–20
These events further increase stress in surgical patients and also alter the function of normal gastrointestinal microflora.3,14,17 Major surgery is associated with severe alterations of the host-defense mechanisms.21–23 In humans who underwent partial gastrectomy, surgical stress rapidly decreased expression of monocyte membrane CD14 and human leukocyte antigen–DR, compared with preanesthesia values.21 Membrane CD14 plays a key role in intracellular transmission of lipopolysaccharide signals and ultimately activating production of tumor necrosis factor-D.22 It has been suggested21,24 that loss of cell surface human leukocyte antigen–DR can reduce the antigen-presenting capacity of monocytes, which results in impaired T-lymphocyte stimulation. These alterations suppress the innate immune system during the perioperative period.21,24 Other studies21,25,26 on surgical stress in humans have revealed that an impaired immune system increases the risk of developing systemic inflammatory response syndrome, sepsis, and multiple organ failure. Studies23,27 in mice have illustrated that surgical stress attributable to laparotomy causes substantial impairment of cell-mediated immunity. In view of the effects of abdominal surgery on the immune system in humans and mice, it is possible that surgical stress similarly suppresses the innate and adaptive immune systems of equine surgical patients. As a result, this may increase their susceptibility to nosocomial Salmonella infections.
In our study, it was difficult to assess the potential interaction effect of abdominal surgery and antimicrobial drug use on nosocomial infection. Assessment of this interaction effect requires the risk estimation of 4 groups of horses: horses unexposed to abdominal surgery and antimicrobial drugs; horses exposed to abdominal surgery only; horses exposed to antimicrobial drugs only; and horses exposed to both factors. In our study, all horses that underwent abdominal surgery were treated with antimicrobial drugs. The expected combined effect of abdominal surgery and antimicrobial drugs was difficult to estimate because our study population did not include a group of horses exposed to abdominal surgery only. We examined a second model that used an indicator variable with 3 categories to estimate the risk of infection in horses exposed to antimicrobial drugs and horses exposed to both abdominal surgery and antimicrobial drugs, compared with the risk for horses not exposed to both factors. The odds of infection were 5 and 15 times as high in horses exposed to antimicrobial drugs and horses exposed to both abdominal surgery and antimicrobial drugs, respectively, compared with the risk for horses not exposed to both factors. In this study, if the odds of infection attributable to abdominal surgery alone were 2, then the estimated OR of 15 as a result of exposure to both factors would be an indication that there was an interaction effect; this would imply that the combined effect of abdominal surgery and antimicrobial drugs on nosocomial infections exceeded that for the additive (eg, 2 + 5 = 7) or multiplicative (2 × 5 = 10) effect of these 2 factors.
The association between antimicrobial drug use and nosocomial Salmonella infections in hospitalized horses has been established in other studies.3,5,6 The normal intestinal flora is an important line of defense against colonization by potentially pathogenic bacteria that are ingested.28 In addition, it averts overgrowth of opportunistic microorganisms already present in the intestines.29,30 Normal intestinal flora prevent the attachment and multiplication of pathogenic microorganisms on mucosal surfaces and their invasion into epithelial cells and the circulation.28 Administration of antibacterial agents, therapeutically or prophylactically, causes disturbances in the ecologic balance between the host and its normal microflora.31 Antimicrobial drugs eliminate intestinal flora that are antagonistic to Salmonella organisms; this may explain the association between exposure to antimicrobial drugs and nosocomial Salmonella infections in horses that are susceptible and previously have had negative results when tested for Salmonella organisms.32 Adverse changes in the function of intestinal tract flora as a result of antimicrobial drug use appear to make surgical patients more susceptible to colonization with Salmonella organisms.
Although frequency of patient–care personnel contacts during the first 3 days after admission was higher in horses that underwent abdominal surgery than in horses that did not, we did not identify an interaction effect between abdominal surgery and number of patient–care personnel contacts on risk of nosocomial infection. This finding indicated that horses that undergo abdominal surgery are at high risk of nosocomial infection and that this risk is not increased nor confounded by a high frequency of patient–care personnel contacts observed in surgical cases.
High caseload alone or in combination with abdominal surgery was not identified as a risk factor for nosocomial Salmonella infection in hospitalized horses. This result can be explained, in part, by the fact that the number of horses shedding Salmonella spp at the time of admission during periods of high or low caseload was not different. Another explanation is that the surveillance and infection control program at the University of Florida large animal hospital is accepted by hospital personnel and the degree of compliance is high. It is possible that high caseload may have an effect on risk of nosocomial infection when the number of shedders in the hospital is high and the size of the veterinary technician task force is not adequate.
Two Salmonella Reading isolates recovered initially from an equine patient and subsequently from the stall used by that horse had different antimicrobial resistance patterns. The first isolate (obtained from the horse) was resistant to amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, and tetracycline, whereas the second isolate (obtained from the stall) was resistant to ceftiofur in addition to amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, and tetracycline. In a recent study,33 Salmonella enterica serotype Infantis isolates from a veterinary teaching hospital had different patterns after restriction enzyme digestion (Xba1 and Spe1) of bacterial DNA, despite the fact that some isolates were recovered from hospitalized horses within the same month and some from the same animal at different time points. In addition, some isolates determined to be similar by pulse-field gel electrophoresis analysis had different susceptibility patterns.33 This variability was attributed to mutations and exchange of genetic material between bacteria throughout the time Salmonella Infantis persisted in the environment, which perhaps was related to selection pressures imposed by the use of antimicrobial drugs and disinfectants in the hospital.33–35 In the study reported here, pulse-field gel electrophoresis analysis was not part of the diagnostic testing. In epidemiologic studies, the use of antimicrobial resistance patterns when matching nosocomial cases to primary cases or environmental isolates can lead to misclassification bias. It is possible that equine patients that shared the same serotype but had a different resistance pattern from that of a primary case horse or an environmental sample (eg, attributable to exchange of resistance plasmids) may have been ruled out as nosocomial case horses. In our study, this type of misclassification was reduced by the surveillance system in place at the hospital, which included the monitoring of data from routine monthly environmental samples as well as that from samples collected from stalls used by equine patients that had positive results for Salmonella spp. For example, in this study, a Salmonella Reading isolate was recovered from a nosocomial case horse; the isolate had a resistance pattern (amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, tetracycline, and ceftiofur) similar to that for an isolate recovered from a stall that had been used by another horse that had positive results for Salmonella Reading but with a different resistance pattern (amoxicillin–potassium clavulanate, ampicillin, chloramphenicol, and tetracycline). Because there was no overlap between the admission and discharge dates for these 2 horses, our definition of nosocomial infection (lacking a positive result for Salmonella Reading in the stall used by the first horse) would not have included the second horse as a nosocomial case horse. Alternatively, because resistance to ceftiofur was tested with a disk diffusion test and given that the minimum inhibitory concentration results were identical, it is possible that this isolate may have been misclassified as resistant to ceftiofur. Because results of the disk diffusion test require visual examination by laboratory personnel, a difference of a few millimeters in the zone of diameter could have caused a potential misclassification (ie, resistance to ceftiofur). Finally, from a risk management point of view for the hospital, because Salmonella Reading isolates were clustered temporally, enhanced infection control measures were maintained until hospital surveillance data revealed no evidence of further disease transmission or contamination attributable to Salmonella Reading.
Results from the epidemiologic study reported here indicated that abdominal surgery increased the risk of nosocomial Salmonella infections in hospitalized horses with gastrointestinal tract disease. In light of this finding, horses that undergo abdominal surgery require enhanced infection control and preventative care. Risk of nosocomial Salmonella infections may be reduced by implementation of preventative measures immediately after surgery, such as housing in an isolation unit and adhering to isolation protocols (eg, use of gloves, gowns, plastic boots, and footbaths). In addition, because of the adverse effects of antimicrobial drugs on the intestinal microflora, justification of antimicrobial use in surgical patients should be carefully evaluated prior to prescription. Finally, this study highlights the importance of environmental sampling for detection of nosocomial cases and to assess the magnitude of disease transmission and hospital contamination as well as the efficacy of cleaning and disinfection procedures.
Abbreviations
CI | Confidence interval |
OR | Odds ratio |
References
- 1.↑
Schott HC II, Ewart SL, Walker RD, et al. An outbreak of salmonellosis among horses at a veterinary teaching hospital. J Am Vet Med Assoc 2001;218:1152–1159.
- 2.↑
Kim LM, Morley PS, Traub-Dargatz JL, et al. Factors associated with Salmonella shedding among equine colic patients at a veterinary teaching hospital. J Am Vet Med Assoc 2001;218:740–748.
- 3.↑
House JK, Mainar-Jaime RC, Smith BP, et al. Risk factors for nosocomial Salmonella infection among hospitalized horses. J Am Vet Med Assoc 1999;214:1511–1516.
- 4.
Ernst NS, Hernandez JA, MacKay RJ, et al. Risk factors associated with fecal Salmonella shedding among hospitalized horse with signs of gastrointestinal tract disease. J Am Vet Med Assoc 2004;225:275–281.
- 5.
Hird DW, Pappaioanou M, Smith BP. Case-control study of risk factors associated with isolation of Salmonella saintpaul in hospitalized horses. Am J Epidemiol 1984;120:852–864.
- 6.
Hird DW, Casebolt DB, Carter JD, et al. Risk factors for salmonellosis in hospitalized horses. J Am Vet Med Assoc 1986;188:173–177.
- 7.↑
National Committee for Clinical Laboratory Standards. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. Approved standard M7–A6. Wayne, Pa: National Committee for Clinical Laboratory Standards, 2003.
- 8.
National Committee for Clinical Laboratory Standards. Performance standards for antimicrobial susceptibility testing. NCCLS document M100–S13. Wayne, Pa: National Committee for Clinical Laboratory Standards, 2003.
- 9.↑
National Committee for Clinical Laboratory Standards. Performance standards for antimicrobial disk and dilution susceptibility tests for bacteria isolated from animals. NCCLS document M31–A2. Wayne, Pa: National Committee for Clinical Laboratory Standards, 2002.
- 10.↑
Hosmer DW, Lemeshow S. Logistic regression for matched casecontrol studies. In: Hosmer DW, Lemeshow S, eds. Applied logistic regression. New York: Wiley, 1989;187–215.
- 11.↑
Daniel WW. Nonparametric and distribution-free statistics. In: Daniel WW, ed. Biostatistics: a foundation for analysis in the health sciences. 3rd ed. New York: John Wiley & Sons, 1983;398–405.
- 12.↑
Hyatt DR, Weese JS. Salmonella culture: sampling procedures and laboratory techniques. Vet Clin North Am Equine Pract 2004;20:577–585.
- 13.↑
Funk A, Peter AD, Monica AN. The effect of fecal sample weight on detection of Salmonella enterica in swine feces. J Vet Diagn Invest 2000;12:412–418.
- 14.↑
Smith BP, Reina-Guerra M, Hardy AJ. Prevalence and epizootiology of equine salmonellosis. J Am Vet Med Assoc 1978;172:353–356.
- 15.
Smith BP. Salmonellosis. In: Smith BP, ed. Large animal internal medicine. St Louis: CV Mosby Co, 1991;818–822.
- 17.
Owen RA, Fullerton J, Barnum DA. Effects of transportation, surgery, and antibiotic therapy in ponies infected with Salmonella. Am J Vet Res 1983;44:46–50.
- 18.
Argenzio RA. Functions of the equine large intestine and their interrelationship in disease. Cornell Vet 1975;65:303–330.
- 19.
Clarke LL, Roberts MC, Argenzio RA. Feeding and digestive problems in horses. Physiologic responses to a concentrated meal. Vet Clin North Am Equine Pract 1990;6:433–450.
- 20.
Linerode PA, Goode RL. The effects of colic on the microbial activity of the equine large intestine, in Proceedings. 16th Annu Conv Am Assoc Equine Pract 1970;321–341.
- 21.↑
Kawasaki T, Ogata M, Kawasaki C, et al. Surgical stress induces endotoxin hyporesponsiveness and an early decrease of monocyte mCD14 and HLA-DR expression during surgery. Anesth Analg 2001;92:1322–1326.
- 22.↑
Ziegler-Heitbrock HW, Ulevitch RJ. CD14: cell surface receptor and differentiation marker. Immunol Today 1993;14:121–125.
- 23.
Zellweger R, Ayala A, Zhu X, et al. Effect of surgical trauma on splenocyte and peritoneal macrophage immune function. J Trauma 1995;39:645–650.
- 24.
Kawasaki T, Ogata M, Kawasaki C, et al. Effects of epidural anaesthesia on surgical stress-induced immunosuppression during upper abdominal surgery. Br J Anaesth 2007;98:196–203.
- 25.
Astiz M, Saha D, Lustbader D, et al. Monocyte response to bacterial toxins, expression of cell surface receptors, and release of anti-inflammatory cytokines during sepsis. J Lab Clin Med 1996;128:594–600.
- 26.
Cheadle WG, Hershman MJ, Wellhausen SR, et al. HLA-DR antigen expression on peripheral blood monocytes correlates with surgical infection. Am J Surg 1991;161:639–645.
- 27.
Qui-shi L, Guizhen Y. Influence of surgical stress on cellular immunity and the induction of plastic adherent suppressor cells of spleen in mice. Immunol Invest 1986;15:419–430.
- 28.↑
Tlaskalova-Hogenova H, Yuekova L, Lodinova-Zadnikova R, et al. Mucosal immunity: its role in defense and allergy. Int Arch Allergy Immunol 2002;128:77–89.
- 29.
Koterba A, Torchia J, Silverthorne C, et al. Nosocomial infections and bacterial antibiotic resistance in a university equine hospital. J Am Vet Med Assoc 1986;189:185–191.
- 30.
Lolekha S. Consequences of treatment of gastrointestinal infections. Scand J Infect Dis Suppl 1986;49:154–159.
- 31.↑
Sullivan A, Edlund C, Nord CE. Effect of antimicrobial agents on the ecological balance of human microflora. Lancet Infect Dis 2001;1:101–114.
- 32.↑
Huntges DJ. Resistance of the indigenous intestinal flora to the establishment of invading microbial populations. In: Schlessinger D, ed. Microbiology. Washington, DC: American Society for Microbiology, 1975;116–119.
- 33.↑
Dunowska M, Morley PS, Traub-Dargatz JL, et al. Comparison of Salmonella enterica serotype Infantis isolates from a veterinary teaching hospital. J Appl Microbiol 2007;102:1527–1536.
- 34.
Graffunder EM, Preston KE, Evans AM, et al. Risk factors associated with extended-spectrum E-lactamase-producing organisms at a tertiary care hospital. J Antimicrob Chemother 2005;56:139–145.
- 35.
Alyaseen SA, Piper KE, Rouse MS, et al. Selection of cross-resistance following exposure of Pseudomonas aeruginosa clinical isolates to ciprofloxacin or cefepime. Antimicrob Agents Chemother 2005;49:2543–2545.
Hydrasponge, International BioProducts, Redmond, Wash.
Hardy Diagnostics, Santa Maria, Calif.
Hardy Diagnostics, Santa Maria, Calif.
API-20E system, Biomerieux Sa, Marcy I'Etoile, France.
Bacto-Salmonella antisera, DIFCO Laboratories, Detroit, Mich.
NMIC30 plate, MicroScan Dade Behring, Siemens Medical Solutions Diagnostics, Deerfield, Ill.
Research Randomizer, Social Psychology Network, Middletown, Conn. Available at: www.randomizer.org. Accessed Jan 10, 2007.
EGRET for Windows, 2000, Cytel Software Corp, Cambridge, Mass.
Statistix for Windows, 2003, Analytical Software, Tallahassee, Fla.