Materials and Methods

Case selection criteria—Data from medical records of all neonatal foals admitted to the Hofmann Equine Neonatal Intensive Care Unit at the University of Florida from January 1982 through June 2007 were entered into a computerized database.^a This database was then reviewed for clinical diagnoses of bacteremia. For the purposes of the study, cultures of blood performed at any time during hospitalization of a foal that yielded bacterial or fungal growth were considered indicative of bacteremia. No other selection criteria were used.

Medical records review—Information obtained from the foals' medical records included breed; sex; year of admission; age at admission; rectal temperature, respiratory rate, and heart rate at admission; sepsis score¹⁶; and clinicopathologic data obtained at admission (total WBC count; total RBC count; PCV; counts of myelocytes, metamyelocytes, band neutrophils, neutrophils, and lymphocytes; plasma protein and fibrinogen concentrations; serum sodium, potassium, chloride, creatinine, total calcium, and glucose concentrations; arterial blood pH; PaCO₂; PaO₂; arterial bicarbonate concentration; and arterial base excess). The names of the dams of all Thoroughbred foals were recorded. Other information included development of localized sepsis-related signs during hospitalization (eg, diarrhea, osteomyelitis, septic arthritis, omphalitis, pneumonia, and meningitis), duration of gestation, duration of hospitalization, outcome (survival to discharge from the hospital or nonsurvival), and concurrent illnesses (eg, hypoxic ischemic encephalopathy, neonatal isoerythrolysis, corneal ulceration, and meconium impaction). Information was also obtained regarding the organisms isolated from blood samples, including identification, antimicrobial susceptibility of identified organisms, and number of isolates if more than 1 species of organisms was identified.

Antimicrobial susceptibility data for penicillin, ampicillin, trimethoprim-sulfamethoxazole, amikacin, gentamicin, tetracycline, chloramphenicol, ceftazidime, ceftiofur, enrofloxacin, imipenem, and cefazolin were collected. The agents selected were those that were consistently included on the antimicrobial susceptibility panel during the entire study period. From January 1982 through December 1992, susceptibility testing was performed by use of the disk susceptibility method. From January 1993 through April 2003, MIC was determined by use of microtitration strips.^17,18,b After April 2003, in vitro susceptibility testing was performed by use of an automated system.^c Regardless of the method, control bacterial strains used to validate the assays at monthly intervals included Staphylococcus aureus (ATCC 29,213), Enterococcus faecalis (ATCC 29,212), Pseudomonas aeruginosa (ATCC 27,853), and E coli (ATCC 25,922 and ATCC 35,218). In all instances, zones of inhibitions or MIC values obtained with the control stains were within the reference range proposed by the Clinical and Laboratory Standard Institute.^19,20 Criteria for interpretation of zones of inhibition or MIC values as susceptible, intermediate, or resistant were also established according to the Clinical and Laboratory Standard Institute guidelines.

For surviving Thoroughbred foals, information regarding racing performance was obtained for each foal and as many as 3 maternal siblings (2 next-oldest and the next-youngest sibling) from a computerized database.^d Only those foals and siblings born during or before 2003 were included because racing data were only available for the years 1950 through 2006; hence, horses were allowed to reach at least 3-year-old status to begin their racing career. Collected data included number of race starts, number of wins, career earnings, and SSI. The SSI is a numeric racing index that is based on the mean earnings per start for Thoroughbred racehorses in North America. This index accounts for year of birth and sex, thereby correcting for inflation and the lower earning potential of fillies. Standard starts index values range from 0.00 to 32.19; an adequate racehorse has an SSI of 1.00, and a superior racehorse is considered to have an SSI of ≥ 3.64, which would place it in the top 3% of all Thoroughbred racehorses.

Data analysis—Descriptive data were compiled and reported as a percentage of the total value, as mean ± SD (if normally distributed), or as median and range (if non-normally distributed). For analysis of factors associated with survival, no differentiation was made between foals that died and those that were euthanized, regardless of reason. Possible factors associated with survival were initially evaluated for tolerance and variance inflation factor (inverse of tolerance). Factors with tolerance < 0.40 and a variance inflation factor > 2.50 were deleted to avoid multicollinearity. Multivariate logistic regression analysis^e was performed with a start model that included all of the following factors: breed (Thoroughbred or non-Thoroughbred); sex; year of admission; age at admission; diarrhea; osteomyelitis; septic arthritis; omphalitis; pneumonia; number of isolates; rectal temperature; respiratory rate; heart rate; sepsis score; PCV; counts of neutrophils, band neutrophils, and lymphocytes; plasma concentrations of protein and fibrinogen; serum concentrations of sodium, potassium, chloride, creatinine, calcium, and glucose; arterial blood pH; PaO₂; and organism classification (yes or no for each of the following: enteric gram-negative, nonenteric gram-negative, gram-positive, and anaerobic organisms and yeast or other fungus). All numeric data were analyzed as continuous variables. Factors were then stepwise deleted, with the factor with the highest P value deleted first. Factors were deleted if the value of P was > 0.30. Odds ratios were then calculated for factors that were significantly (P < 0.05) associated with survival. Odds ratios are presented as the point estimate and 95% CI. An OR > 1 corresponds to a positive association, whereas a ratio < 1 corresponds to a negative association. For example, a categoric variable with an OR of 5 means that a foal with that variable is 5 times as likely to survive as a foal without that variable. For every incremental increase in a continuous variable with an OR of 5, a foal with that incremental increase would be 5 times as likely to survive as a foal without that incremental increase. Factors associated with localized infections (diarrhea, septic arthritis, osteomyelitis, omphalitis, or pneumonia) were assessed via multivariate logistic regression in the manner described for survival analyses.

Data regarding bacterial isolates were grouped according to the decade in which they were collected (ie, 1982 through 1989, 1990 through 1999, and 2000 through 2007) and by organism classification (enteric gram-negative, nonenteric gram-negative, and grampositive organisms). Anaerobes and yeast or other fungus were excluded because of the small number of isolates. Organisms for which at least 9 isolates were obtained during the entire study were also evaluated individually. Comparisons of decade groupings and individual organisms or classes of organism were made via χ² analysis.^f When significant differences were detected, post hoc F² analysis was performed to compare each organism with all others among the decade groupings. The value of P that was considered significant was 0.05/m, where m is the number of comparisons via Bonferroni correction. Factors associated with organism classification were evaluated via multivariate logistic regression, in the manner described for survival analyses.

Antimicrobial susceptibility data were analyzed for all organisms, organisms grouped by decade (1982 through 1989, 1990 through 1999, and 2000 through 2006), and organisms grouped by classification (enteric gram-negative, nonenteric gram-negative, and grampositive organisms). On the basis of the antimicrobial susceptibility data, organisms were classified as susceptible or nonsusceptible (ie, intermediate or resistant) to each antimicrobial agent. Comparisons by decade grouping were made via F² analysis if the number of data sets was > 5 or via minimum F² analysis if the number of data sets was ≤ 5.^f A value P = 0.05/m was considered significant, where m is the number of comparisons via Bonferroni correction.

For analysis of racing performance of foals with bacteremia that survived, year of birth, percentage of starters, percentage of winners, number of wins, number of starts, earnings, and SSI values were compared with the mean values for all siblings by use of a Wilcoxon signed rank test.^f A value of P ≤ 0.05 was considered significant.

Results

Of 2,188 foals admitted to the hospital during the study period, 423 foals met the inclusion criteria; 254 (60%) of those foals survived to discharge from the hospital. Among the 423 foals, there were 198 fillies and 219 colts (6 foals did not have sex recorded). Most foals were Thoroughbreds (n = 269), 34 were Quarter horses, 23 were Arabians, 18 were Warmbloods, 13 were American Miniature horses, 11 were mixed breeds, 10 were Paso Finos, 9 were Appaloosas, 5 were Tennessee Walking horses, 5 were Paints, 3 were Standardbreds, and 23 were other breeds or had no breed recorded. For all analyses, only significant effects are discussed.

Outcome—For short-term survival analyses, only 250 foals (148 survivors) had complete data for all factors. The following factors were negatively associated with survival in bacteremic foals: age at admission, septic arthritis, band neutrophil count, and serum creatinine concentration (Table 1). Factors positively associated with survival included year of admission, diarrhea, rectal temperature, neutrophil count, and arterial blood pH.

Localized infections and concurrent disease processes—The numbers of survivors and nonsurvivors that had a concurrent disease process or localized infection were assessed (Table 2). Factors that were positively associated with diarrhea included enteric gramnegative bacteria (OR, 5.85; 95% CI, 2.08 to 16.45), year of admission (OR, 1.10; 95% CI, 1.01 to 1.19), age at admission (OR, 1.01; 95% CI, 1.00 to 1.02), omphalitis (OR, 15.29; 95% CI, 2.99 to 78.25), serum chloride concentration (OR, 1.11; 95% CI, 1.03 to 1.19), and anaerobic bacteria (OR, 80.72; 95% CI, 2.75 to > 999.99). Factors negatively associated with diarrhea were breed (the odds of a foal developing diarrhea were 1.00 to 0.29, or 71% less in breeds other than Thoroughbreds), pneumonia (OR, 0.28; 95% CI, 0.08 to 0.94), serum total calcium concentration (OR, 0.57; 95% CI, 0.40 to 0.82), and arterial blood pH (OR, 0.63; 95% CI, 0.41 to 0.97).

Factors that were positively associated with osteomyelitis included septic arthritis (OR, 8.87; 95% CI, 2.38 to 33.01) and enteric gram-negative organisms (OR, 4.13; 95% CI, 1.01 to 16.92). Factors that were positively associated with septic arthritis included osteomyelitis (OR, 14.50; 95% CI, 3.99 to 52.62) and age at admission (OR, 1.01; 95% CI, 1.003 to 1.01). Diarrhea (OR, 7.766; 95% CI, 1.98 to 30.46), rectal temperature (OR, 1.60; 95% CI, 1.13 to 2.27), sepsis score (OR, 1.15; 95% CI, 1.01 to 1.32), and year of admission (OR, 1.11; 95% CI, 1.02 to 1.22) were positively associated with omphalitis. Packed cell volume (OR, 0.94; 95% CI, 0.88 to 0.998), serum sodium concentration (OR, 0.93; 95% CI, 0.88 to 0.996), enteric gramnegative organisms (OR, 0.229; 95% CI, 0.06 to 0.86), and gram-positive organisms (OR, 0.1; 95% CI, 0.02 to 0.62) were negatively associated with omphalitis. Factors that were negatively associated with pneumonia included year (OR, 0.92; 95% CI, 0.88 to 0.96), diarrhea (OR, 0.34; 95% CI, 0.16 to 0.70), and PaO₂ (OR, 0.99; 95% CI, 0.98 to 1.00). Culture of yeast from blood samples was positively associated with pneumonia (OR, 9.57; 95% CI, 1.57 to 58.19).

Causative organisms—From the 423 foals, 554 separate isolates were obtained. Of these, 527 were obtained at the time of admission and 27 were obtained from cultures of blood samples collected subsequently. Overall and for each decade grouping, E coli was the most frequently isolated organism (Table 3). During the study period, Actinobacillus spp were more commonly isolated in 1982 through 1989 than in later years (1990 through 1999 [P = 0.005] and 2000 through 2007 [P = 0.012]) and Salmonella spp were isolated less frequently in 2000 through 2007 than in previous years (1982 through 1989 [P = 0.016] and 1990 through 1999 [P = 0.01]). Enteric gram-negative organisms were less frequently isolated from blood samples from bacteremic foals in 2000 through 2007, compared with findings in 1982 through 1989 (P < 0.001). No other comparisons revealed significant differences.

Factors that were positively associated with enteric gram-negative organisms included diarrhea (OR, 1.78; 95% CI, 1.08 to 2.92), osteomyelitis (OR, 4.04; 95% CI, 1.03 to 15.88), serum chloride concentration (OR, 1.03; 95% CI, 1.00 to 1.06), and blood glucose concentration (OR, 1.003; 95% CI, 1.00 to 1.01). Number of isolates was negatively associated with enteric, gramnegative organisms (OR, 0.56; 95% CI, 0.36 to 0.88). Diarrhea (OR, 0.04; 95% CI, 0.01 to 0.27) and number of isolates (OR, 0.18; 95% CI, 0.04 to 0.74) were negatively associated with nonenteric gram-negative organisms. Number of isolates (OR, 0.08; 95% CI, 0.02 to 0.34), PCV (OR, 0.96; 95% CI, 0.92 to 1.00), and total lymphocyte count (OR, 0.999; 95% CI, 0.999 to 1.00) were negatively associated and rectal temperature (OR, 1.16; 95% CI, 1.00 to 1.35) was positively associated with gram-positive organisms.

Antimicrobial susceptibility—Antimicrobial susceptibility data obtained from the medical records were analyzed (Table 4). Among the nonenteric gram-negative bacteria, antimicrobial susceptibility patterns did not change during the study period. Enteric gram-negative bacteria became more susceptible to tetracycline and to 1 or both agents in penicillin-amikacin and ampicillinamikacin treatment combinations; however, susceptibility patterns for the other antimicrobials did not change over time. Gram-positive bacteria became more susceptible to chloramphenicol during the study period, but susceptibility patterns for the other antimicrobials did not change over time. All bacteria became more susceptible to tetracycline and chloramphenicol and less susceptible to enrofloxacin over time; changes in susceptibility patterns for the other antimicrobials were not evident.

Racing performance—Racing data were available for 102 of the 126 surviving Thoroughbred foals born prior to 2004. Racing data were available for 194 siblings (74 next-oldest siblings [sibling 1 group], 54 second-oldest siblings [sibling 2 group], and 66 nextyoungest sibling [sibling 3 group]). Median values and upper limit of the range for all data were calculated (Table 5). The year of birth was significantly (P < 0.001) higher for foals than for siblings. There were no significant differences between surviving foals and siblings with regard to percentage of starters, percentage of winners, or number of race starts. Compared with their siblings, surviving foals had significantly lower number of wins, total earnings, and SSI values.

Discussion

The overall proportion of bacteremic foals that survived to discharge from the hospital in the period investigated in the present study was similar to previously reported findings.^8,9,13 Not surprisingly, survival of bacteremic foals improved over time. Some of the factors that were associated with survival in the present study have been identified in other studies, including neutrophil count¹⁰ and rectal temperature.^6,10 The negative association between survival and septic arthritis in our study could be somewhat misleading in that differentiation between death and euthanasia of foals that did not survive was not attempted. The perception of decreased athletic outcome²¹ associated with this type of infection may have influenced the owners' decision not to pursue additional treatment. But because of the retrospective nature of the study, such a differentiation was not possible in all cases and was therefore not attempted. The negative association of band neutrophil count and the positive association of neutrophil count with survival highlight the fact that a severe left shift and neutropenia associated with severe endotoxemia are negative prognostic indicators. A positive association between neutrophil count and survival in a population of critically ill foals has been previously reported.¹⁰ However, the association between diarrhea and survival among bacteremic foals was interesting and, to the authors' knowledge, has not been previously reported, although diarrhea has been associated with Actinobacillus-related bacteremia.⁸ The association between diarrhea and survival may just reflect the fact that, of the localized signs associated with bacteremia, diarrhea is the least detrimental to overall outcome.

In the present study, enteric gram-negative isolates were positively correlated with development of diarrhea and osteomyelitis, but negatively correlated with omphalitis. Development of diarrhea and omphalitis were positively associated with each other, whereas development of diarrhea and pneumonia were negatively correlated with each other. This may reflect overall incidence of infections in our hospital population, possible routes of infection, or that those infections localized in the thorax did not typically cause diarrhea. Not surprisingly, osteomyelitis and septic arthritis were positively correlated with each other. Interestingly, an association between omphalitis and infectious orthopedic conditions was not evident in the current study. Omphalitis was the only localized infection positively correlated with rectal temperature at admission. If the development of fever at any point during hospitalization had been recorded, this association may have been different because many localized infections are not apparent at the time of admission.

During the entire study period and within each decade grouping, E coli was the organism most commonly isolated in cultures of blood samples from bacteremic foals, similar to findings of other investigations.^3–9 Isolation of Actinobacillus spp, Salmonella spp, and enteric gram-negative organisms became less common during the study period. Such changes over time within the same referral population have been infrequently reported. In a report⁶ of a study of septicemic neonatal foals in Georgia, there was a dramatic decrease in the percentage of E coli isolates detected from 1986 through 1990 and in later 5-year sampling periods (1991 through 1995 and 1996 through 2000); during the same periods, the percentage of Enterococcus and Staphylococcus spp detected increased. We had initially hypothesized that a change from a predominance of gram-negative organisms to predominance of gram-positive organisms over time could be partially accounted for by an increase in hospital-acquired organisms. The number of isolates obtained from cultures performed after admission (ie, organisms that had been acquired after admission) was too small to include this association in final analyses. Nevertheless, the relative frequency of isolates acquired after admission to the hospital and those acquired in the field (detected at time of admission) did not appear to differ. Implementation of biosecurity procedures on farms may possibly decrease the overall incidence of gram-negative infections.

Among the microorganisms cultured from blood samples collected from bacteremic foals during the 25.5-year study period, one might expect antimicrobial resistance to have developed; however, such a change was only evident for enrofloxacin. Because changes in laboratory protocols were almost coincidental with changes in decade, it is difficult to determine what effect that additional variable had on antimicrobial susceptibility patterns. However, because laboratory protocols, qualitycontrol procedures, and data interpretation conformed to Clinical and Laboratory Standard Institute guidelines during each period, it is unlikely that the change in methods had a major effect on the percentages of susceptible microorganisms. The increased susceptibility of enteric gram-negative isolates to 1 or both agents in the treatment combinations of penicillin and amikacin and ampicillin and amikacin was not expected because those treatments represented the antimicrobial combinations most frequently used in the study population. Although susceptibility to trimethoprim-sulfamethoxazole did not change over time, the percentage of susceptible organisms in our study was greater than that determined in another investigation.⁷ This likely reflects regional variability in infective organisms and their antimicrobial susceptibility patterns.

To the authors' knowledge, this is the first large-scale evaluation of racing performance in a population of surviving bacteremic foals. Although surviving foals in our study were just as likely to race or win as their siblings, they did not earn as much money and had a lower overall performance, as indicated by the SSI rating. These findings are somewhat in contrast to those reported for surviving Thoroughbred and Standardbred foals from neonatal intensive care units (admitted because of bacteremia and other reasons), in which the percentage of starters was lower than the percentage in the control population (the foals' dams and siblings); however, on comparison of those 2 groups, performance over a 2-year period was not different in horses that were able to make at least 2 race starts.¹⁵ Many factors influence athletic performance; therefore, maternal siblings were used as the control population in the present study in an effort to minimize genetic variability. But because at least 1 race start by a horse implies some degree of athletic potential, one cannot fully explain whether the decreased performance was truly a lingering effect of neonatal disease or was related to other factors.