Differences in isolation rate and antimicrobial susceptibility of bacteria isolated from foals with sepsis at admission and after ≥48 hours of hospitalization

Abstract Background Antimicrobial treatment protocols for foals with sepsis that do not improve clinically often are adjusted based on bacteriological and antimicrobial susceptibility testing results from samples collected at hospital admission. Objectives To evaluate whether hospitalization for ≥48 hours affects bacteriological and antimicrobial susceptibility testing results. Animals Two‐hundred sixty‐seven foals <30 days of age admitted to a neonatal intensive care unit and diagnosed with sepsis. Methods Medical records were reviewed retrospectively to identify foals with sepsis and positive bacteriological cultures. Results from samples collected at hospital admission were compared to those collected ≥48 hours after admission. Logistic regression for clustered data and exact logistic regression were used for statistical analysis. Results Three‐hundred fifty‐three unique bacterial isolates were obtained from 231 foals at hospital admission and 92 unique bacterial isolates were obtained from 57 foals after ≥48 hours of hospitalization. Relative isolation frequency after ≥48 hours of hospitalization increased for Acinetobacter spp., 0.6% versus 3.3% (odds ratio [OR], 7.63; 95% confidence interval [CI], 1.28‐45.45); Enterococcus spp., 4.8% versus 19.6% (OR, 5.37; 95% CI, 2.64‐10.90); Klebsiella spp., 5.1% versus 10.9% (OR, 2.27; 95% CI, 1.05‐4.89); Pseudomonas spp., 3.0% versus 7.6% (OR, 3.49; 95% CI, 3.49‐240.50); and Serratia spp., 3.0% versus 5.4% (OR, 20.23; 95% CI, 2.20‐186.14). Bacteria isolated after ≥48 hours of hospitalization were less susceptible to all tested antimicrobial drugs, except for imipenem. Conclusions and Clinical Importance Decreased antimicrobial susceptibility of bacteria isolated after ≥48 hours of hospitalization provides a rationale for repeated bacteriological culture and susceptibility testing in hospitalized foals with sepsis.


| INTRODUCTION
Up to 60% of foals admitted to an intensive care unit in Florida were considered septic at hospital admission. 1 Escherichia coli is the bacterium most commonly isolated from foals with sepsis in most studies. [2][3][4][5][6][7][8] Antimicrobial susceptibility of bacteria isolated from foals with sepsis varies among different geographic regions. 3,5,6,[8][9][10][11][12] Temporal trends toward increased antimicrobial resistance to frequently used antimicrobial drugs, such as gentamicin, amikacin, and ceftiofur, have been identified. 12 This finding highlights the need to perform bacteriological culture and susceptibility testing in foals suspected of sepsis.
Bacteriological and antimicrobial susceptibility testing is performed routinely on samples collected from foals with suspected sepsis at hospital admission. 13 While awaiting test results, the choice of antimicrobials to initiate treatment typically is based on historical data on antimicrobial susceptibility patterns of pathogens causing sepsis of foals in that geographic location. 3,5,6,[8][9][10][11][12] The antimicrobial treatment regimen then is adjusted as necessary based on the results of culture and susceptibility testing of admission samples. Although this approach results in a successful outcome in 65% of affected foals, 35% fail to show clinical improvement despite treatment with antimicrobials that should be effective based on susceptibility testing of bacteria isolated from admission samples. 13 Clinicians then may opt to give the chosen antimicrobial protocol more time to be effective or adjust the treatment protocol to include other antimicrobials to which the bacteria isolated from admission samples were susceptible. Both of these approaches assume that the bacterial species infecting the foal and the antimicrobial susceptibility profile of these bacteria remain the same as those obtained from admission samples.
In adult horses, hospitalization and treatment with antimicrobial drugs create selection pressure on bacteria, leading to the development of antimicrobial resistance. [14][15][16][17] Several studies have reported on isolation rate and susceptibility patterns of bacteria isolated from foals with sepsis. 2,3,5,6,[8][9][10][11][12] However, the effect of hospitalization and antimicrobial treatment before the time of sampling on culture results has not been investigated in foals with sepsis.
The Centers for Disease Control and Prevention (CDC) defines health care-associated infections (HAIs) as localized or systemic conditions resulting from an adverse reaction to the presence of an infectious agent(s) or its toxins in which case there is no evidence that the infection was present or incubating at the time of hospital admission. 18 Critically illhuman patients admitted to intensive care units are at risk of developing HAIs, frequently related to particular surgical and medical procedures, which often involve specific species or strains of bacteria that are resistant to many antimicrobial drugs. 19 The same is likely true for foals admitted to neonatal intensive care facilities, but data to support this assumption currently are lacking. The main purpose of our study was to compare isolation rates and antimicrobial susceptibility patterns of bacteria isolated from foals with sepsis between samples collected on hospital admission and after ≥48 hours of hospitalization. The 2nd aim was to determine if HAIs occurred in foals after ≥48 hours of hospitalization.
The hypotheses were that different bacterial species would be isolated after ≥48 hours of hospitalization compared to samples collected at hospital admission and that these bacteria would be more resistant to antimicrobials. Also, we hypothesized that a large proportion of positive bacterial cultures after ≥48 hours of hospitalization potentially would be the result of HAIs.
were kept refrigerated after death or euthanasia until the necropsy was performed on the day of death or the next day. Cases were included only if foals showed ≥5 clinical or pathologic signs of systemic sepsis at the time of sample collection, such as fever (>38.9 C), neutropenia, or neutrophilia (<4000 or > 12 000 neutrophils/μL), increased band neu-

| Bacterial isolation, identification, classification, and antimicrobial susceptibility testing
Bacterial isolation, identification, and classification were performed as described previously. 12 The broth microdilution Sensititre procedure (ThermoFisher Scientific, Cleveland, Ohio) was used for antimicrobial susceptibility testing, following Clinical Laboratory Standards Institute (CLSI) protocols. 20 The minimum inhibitory concentration (MIC) was recorded as the lowest concentration of antimicrobial drug that inhibited visible growth of bacteria or 80% inhibition in the case of trimethoprim/sulfamethoxazole (TMS). Breakpoints published in the 3rd edition of the "Performance Standards for Antimicrobial Disk and Dilution Susceptibility Tests for Bacteria Isolated From Animals" by CLSI were used to determine susceptibility for all isolates included in the study, occasionally modified based on research in horses (see Table S1). 20 For foals that were treated with combinations of antimicrobial drugs, an isolate was considered to be susceptible to this combination of drugs if its MIC for at least 1 of the drugs in the combination was equal to or less than the breakpoint.

| Antimicrobial drugs
The following antimicrobial drugs were evaluated for activity against bacteria isolated from foals with sepsis: amikacin, ampicillin, ceftiofur, chloramphenicol, enrofloxacin, gentamicin, imipenem, penicillin, tetracycline, and TMS. Based on the measured antimicrobial activity of these drugs, the susceptibility of individual isolates to the following combinations of drugs, which frequently are used to treat foals with sepsis, was predicted: amikacin + penicillin, amikacin + ampicillin, amikacin + ceftiofur, gentamicin + penicillin, and gentamicin + ampicillin. isolates that also were obtained at both time points, but were, on both occasions, resistant to the initial antimicrobial treatment.

| Statistical analysis
Logistic regression using cluster robust SE estimation was used to assess the association between isolation frequency and antimicrobial susceptibility of bacteria isolated at hospital admission as compared to those isolated after ≥48 hours of hospitalization (Stata/IC 14.1). 21,22 Several potential confounders of these differences were identified before analysis of the data: "antemortem versus postmortem culture," "culture site," "year of culture," and for the detection of differences in susceptibility patterns, "bacterial species isolated" also were identified.
The potential confounding effects of these variables were assessed by comparing the results of the single-variable analysis to those of the multivariable analysis, including the potential confounder as covariate.
A change of ≥10% in the odds ratio (OR) was considered evidence of sufficient confounding to justify retention of the variable in the model regardless of its statistical significance; otherwise, the variable was excluded from the model. Because a strong association was found between time of sampling ("hospital admission" versus "after ≥48 hours of hospitalization") and the variables "antemortem versus postmortem culture" and "culture site" (eg, postmortem cultures were overrepresented in the group of "samples collected after ≥48 hours of hospitalization" and blood cultures were overrepresented in the group of "samples collected at hospital admission"), the method described above could not be applied to "antemortem versus postmortem culture" and "culture site." Therefore, a chi-square test was used to assess the association of these variables with isolation frequency and antimicrobial susceptibility within the group of "samples collected at hospital admission" only. Because they were not found to be significantly associated with either isolation frequency or antimicrobial susceptibility, both variables were excluded from further analyses.
Inclusion or exclusion of the covariates is shown in Tables 1-3.
When some of the cells formed by the outcome and predictor variable had no observations, exact logistic regression was used instead of ordinary logistic regression; correction for within-cluster correlation was maintained. The statistical methods used are noted in the tables presenting the results (see Tables 1-3). Associations are expressed as ORs with 95% confidence intervals (CIs). Statistical significance was defined as P < .05.

| RESULTS
A total of 445 bacterial isolates from 267 foals were included in this study. Three-hundred fifty-three isolates were obtained from samples The majority of these bacteria were isolated from various organs at necropsy (n = 46), but bacteria also were isolated from blood cultures (n = 11), joint aspirates (n = 11), peritoneal fluid samples (n = 11), and IV catheter tips (n = 13

| Isolation frequency
Escherichia coli was isolated most frequently from samples collected on the day of hospital admission, followed by Actinobacillus spp. and T A B L E 1 Isolation frequency of bacteria cultured from foals with sepsis at admission versus after >48 hours of hospitalization

| Antimicrobial susceptibility
Susceptibility data, specified by antimicrobial drug or combination of antimicrobial drugs, are presented in Escherichia coli was the only bacterial species for which the number of isolates was high enough to make a meaningful comparison between hospital admission and after ≥48 hours of hospitalization.
Thus, we also have included separate susceptibility data for E. coli (   Of the drugs included in our study, ceftiofur and enrofloxacin are classified as "highest priority critically important antimicrobials" by the World Health Organization (WHO), which means they are regarded as critically important to human health. 23 Amikacin, ampicillin, gentamicin, and imipenem are classified as "high priority critically important antimicrobials". Chloramphenicol, penicillin, tetracycline, and TMS are "highly important antimicrobials" according to the WHO. The use of "highest priority critically important antimicrobials" in horses should be restricted, according to the WHO, and should be reserved only for cases for which no alternative antimicrobials are effective, and only after appropriate susceptibility testing. In our opinion, this policy also should be applied to imipenem, because it is the only "high priority critically important antimicrobial" we tested that is listed as the sole treatment available for specific diseases in humans.

| Isolation frequency
After ≥48 hours of hospitalization, the odds of samples being positive for Actinobacillus spp. and Streptococcus spp. significantly decreased compared to samples collected at admission. These bacterial species typically are susceptible to antimicrobial drugs commonly used for initial treatment of foals with sepsis, such as the combination of amikacin and ampicillin. 12 Therefore, it is not surprising that in samples collected after ≥48 hours of hospitalization these bacteria were isolated less frequently.
After ≥48 hours of hospitalization, the odds of bacterial cultures being positive for Acinetobacter spp., Enterococcus spp., Klebsiella spp., Pseudomonas spp., and Serratia spp. all significantly increased. A high proportion of these species of bacteria are known to show intrinsic or acquired resistance to many antimicrobial drugs, including those commonly used in initial treatment protocols for foals with sepsis. [24][25][26][27][28][29] These bacterial species were responsible for a large proportion of the HAIs in a large study in humans: Enterococcus spp., 13.9%; Klebsiella spp., 8%; Pseudomonas spp., 7.5%; Serratia spp., 2.1%; and Acinetobacter spp., 1.8%. 30 In equine medicine, there also are several reports on the role of these bacteria in HAIs. A study on Acinetobacter baumanni isolates from companion animals and horses in Switzerland found that a majority of these infections were hospital-acquired. 31 In a study of surgical site infection after laparotomy in horses, Enterococcus spp. were the 2nd most commonly isolated bacteria. 32 Klebsiella spp. were identified as causative organisms of pneumonia in 11 horses that had undergone mechanical ventilation under general anesthesia. 33 The relatively high proportion of bacterial species that are known to frequently cause HAIs isolated in our study from samples after ≥48 hours of hospitalization suggests that HAIs could also play an important role in equine neonatal care. Further genotypic characterization would be required to confirm this hypothesis.

| Antimicrobial susceptibility
Bacteria isolated after ≥48 hours of hospitalization were less susceptible to all antimicrobial drugs and combinations evaluated in our study compared to those isolated at admission. This decreased susceptibility was significant for all drugs and drug combinations, except for imipenem.
None of the antimicrobial drugs or their combinations were predicted to have an efficacy of >90% against bacteria isolated after ≥48 hours of hospitalization. Susceptibility patterns of these bacteria were unpredictable.
This observation can be explained in part by the different bacterial species that were isolated. Antimicrobial treatment was initiated in all foals included in our study at hospital admission after collection of the 1st sample for bacteriological culture and susceptibility testing. This approach likely led to an antimicrobial selection pressure, favoring growth of resistant bacterial populations, as is also seen in studies in adult horses. [14][15][16][17] Therefore, the bacterial species and strains isolated after ≥48 hours of hospitalization frequently were more resistant to multiple antimicrobial drugs compared to the species and strains isolated at hospital admission.
However, the results for E. coli clearly indicate that, even within the same bacterial species, the odds of being susceptible significantly decreased between admission and ≥48 hours after admission (

| Limitations
We acknowledge that several aspects of the design of our study could have influenced the results obtained. First, only isolates originating from samples from foals treated at the University of California-Davis VMTH were included, which could have led to geographically restricted findings. Second, our study was largely based on a retrospective review of medical records, and therefore cases for which essential information was missing were excluded. Given the retrospective nature of our study, we did not have follow-up samples for bacteriological culture and susceptibility testing available for all foals included in the study. At hospital admission, blood cultures were collected routinely in foals suspected of sepsis, whereas later sampling during hospitalization was based on the clinical situation (eg, poor treatment response). This situation potentially could have created substantial sampling bias and prevented conclusions being drawn regarding potential mechanisms for the observed differences in isolation frequency and antimicrobial susceptibility. However, this factor does not restrict the clinical value of the data in guiding clinicians who need to decide whether to adjust antimicrobial treatment protocols in foals with on-going sepsis. Not all susceptibility testing was performed at the same time, although the same methods were used throughout the study and the same interpretation criteria regarding antimicrobial susceptibility were applied to all isolates included in the study. 20 Administration of antimicrobial drugs before hospitalization could have influenced susceptibility profiles of bacteria isolated at hospital admission. Data on antimicrobial treatment before admission were not available for all cases and could not be taken into account. Antimicrobial removal devices (ARD) were not consistently used for blood cultures throughout the study period. Without use of an ARD, inhibition of growth of susceptible bacteria in vitro may have occurred and given false negative culture results in some cases.
Samples collected postmortem may have a higher risk of contamination compared to samples collected antemortem. To minimize the risk of this factor influencing our results, isolates collected postmortem only were included if they were isolated from at least 2 normally sterile sites.