Effect of prior receipt of antibiotics on the pathogen distribution: a retrospective observational study on 27 792 patients

Background There have been no systematic studies of microbiological differences before and after antibiotics treatment. The aim of this study was to evaluate the effect of prior receipt of antibiotics on the microorganism distribution. Methods A retrospective, observational study patients was conducted in a 3,200-bed tertiary, referral, teaching hospital in eastern China. During a 2-year period, all hospitalized patients treated with antimicrobial agents were enrolled in this study. Among 48,692 patients evaluated, the 27,792 (57.1%) who were sampled within two days before or after administration of the first dose of antimicrobial agents were included. Distribution of clinical specimens and the microorganism were compared between before and after antibiotic drug treatment groups. Results Compared to specimens taken after antibiotics exposure, specimens taken before antibiotics exposure had a higher proportion of blood and urine specimens and a higher culture positive rate (all P <0.001). Higher percentages of Staphylococcus aureus (9.9% vs. 8.5%, P =0.041), non-fermenting bacteria (27.7% vs. 19.9%, P <0.001), and fungi (7.9% vs. 3.6%, P <0.001) were isolated from the group after antibiotics exposure, while the percentages of Streptococcus spp. (4.8% vs. 2.7%, P <0.001), Haemophilus influenzae (2.3% vs. 0.8%, P <0.001), and Moraxella catarrhalis (0.7% vs. 0.1%, P <0.001) were higher in the group before antibiotics exposure. Further analysis found significant differences or different rank orders of microbes derived from respiratory secretions, blood or urine samples, but no statistical difference was found for microbes isolated from body fluid specimens between the two groups.


Background
Antimicrobial resistance has increased in recent decades. Antimicrobial stewardship is important for addressing the problem of resistant pathogens, and obtaining specimens prior to the administration of antimicrobials is a key part of this program. Sterilization of cultures may occur shortly after the use of antibiotics [1][2], which may lead to negative culture results as described in some research [3][4][5][6][7][8][9]. Moreover, studies conducted by Montravers et al. [10] and Harbarth et al. [11] showed that starting antibiotic therapy before sample collection may result in detecting less-sensitive microorganisms. However, the former only included 76 consecutive patients with ventilator-associated pneumonia [10], and the latter [11] mainly evaluated gram-negative pathogens. In addition, one study assessed the impact of prebiopsy antibiotics on pathogen recovery in hematogenous vertebral osteomyelitis patients and revealed that antibiotic exposure before biopsy did not negatively impact pathogen recovery [12]. Therefore, a systematic study of the microbiological differences isolated from samples taken before and after the initiation of antibiotic therapy, which to the best of our knowledge is lacking in the literature, is necessary.
Here, we developed a computerized antimicrobial decision-support system (aCDSS) embedded in our hospital's electronic medical record system (EMRS) that integrated into the clinical workflow the requirement that doctors obtain microbiological specimens before antibiotic therapy starting in 2015. With the assistance of aCDSS, we were able to classify all clinical specimens according to collection time and record the microorganisms that were present, thus analyzing the effects of antimicrobial agents on the distribution of organisms. The overall aim of this study was to assess the potential association between antimicrobial agent therapy and the distribution of microbes isolated from clinical 4 specimens.

Study design and setting
This study was performed at the second affiliated hospital of Zhejiang University School of Medicine, a 3 200 bed tertiary referral and teaching hospital in China. The protocol was approved by human research ethics committee of the second affiliated hospital of Zhejiang University School of Medicine (2018-025). Due to the retrospective observational nature of the study, the Institutional Review Board waived the need for informed consent.
All hospitalized patients treated with antibiotics discharged from 1 January 2015 to 31 December 2016 in our hospital were enrolled. Patients who had clinical specimens collected within 2 days before or after antibiotic therapy were included, while those who did not have clinical specimens collected during this period were excluded. Clinical specimens of infection sources were obtained as clinically indicated, but only the first sample (infection episode) during this period (within 4 days as described above) per patient was included in the analysis. With the assistance of aCDSS, we recorded the accurate time of the first dose of therapeutic antibiotics for each infected (or suspected to be infected) patient and the time that the specimen was taken, as well as the corresponding microorganism results. A polymicrobial result was defined as more than one pathogen cultured from the same specimen. We also took executive infection control measures during this study period in our hospital. All of the patients were treated by the attending physician.

aCDSS
The aCDSS embedded in our hospital's EMRS was developed specifically to target antibiotics, including prophylaxis and therapeutic use, and was linked to the hospital laboratory information system (LIS) in real-time to provide physicians with the latest 5 patient information. When prescribing antibacterial drugs, the system automatically decides whether the purpose is for "therapeutic" or "prophylactic" use according to the repository set up by our research group. For therapeutic use, the aCDSS integrated into the clinical workflow requires doctors to obtain microbiological specimens before starting antibiotic therapy. In addition, nurses need to use a personal digital assistant (PDA) to scan a two-dimensional code of samples before using antimicrobial drugs. Therefore, aCDSS can accurately record the time of antibiotic use as well as the time of sample taken.

Data collection
Cases were assigned randomly to 1 of 3 trained extractors, who extracted demographic data. Subjects were assigned to two groups according to the time sequence of sampling and the first dose of antibiotics: the SBA group (specimen taken before antibiotic therapy) and SAA group (specimen taken after antibiotic therapy).

Statistical analysis
All statistical analyses were conducted using MS Excel 2016. Data are expressed as means ± standard deviation (SD) for continuous variables, and as frequencies (percent) for categorical variables. Groups were compared using Student's t test for continuous variables and the chi-square (c 2 ) test or Fisher's exact test for categorical data as appropriate. All statistical tests were 2-tailed, and P< 0.05 indicated statistical significance.

Results
During the study period, a total of 48 692 patients were treated with various antibiotics, with samples taken from 27 792 (57.1%) patients within 2 days before or after starting antibiotic treatment. Among them, 19 868 (71.5%) of patient samples were taken before 6 antibiotic therapy (SBA group), and 7 924 (28.5%) of patient samples were taken after antibiotic therapy (SAA group).

Patient characteristics and culture positive rate
The demographic characteristics of the included patients are shown in Table 1 d Polymicrobial result was defined as more than one pathogen cultured from the same specimen.

Clinical specimen distribution
There was a significant difference in clinical specimen distribution between the two groups as shown in Table 1. The proportion of blood and urine specimens taken from the SBA group was higher than that of the SAA group (

Microbial distribution isolated from specimens
Final culture results were made available by aCDSS, which yielded a total of 8 850 isolates, with 6 516 strains isolated from the SBA group and 2 334 from the SAA group.
Differences in microbial distribution between the two groups are shown in Table 2. 0.1% in the SAA group, respectively, both P<0.001).

Fungi
The isolation rate of Candida spp. in the SAA group was 7.9%, which was higher than that in the SBA group (3.6%, P<0.001). None of the other species had significant differences between the two groups.

Subgroup analysis
As a result of inconsistent sample composition between the two groups, which may lead to bias, we conducted subgroup analysis to compare the microorganism distribution according to the specimen types.

Respiratory secretions
Further analysis of respiratory secretions found that the microbial ratio was significantly different between the two groups. Fig 1 shows the

Discussion
The main finding of this study is that the distribution of microorganisms isolated from specimens taken before and after antibiotic treatment was different, and in most cases, sensitive organisms were more easily isolated before antibiotic therapy, while druginsensitive organisms such as non-fermentative bacteria and fungi, were more frequently isolated after antibiotic exposure. However, the effect of prior receipt of antibiotics on the pathogen distribution is specimen-dependent and this trend was not obvious in body fluid specimens.
Studies conducted by Montravers et al. [10] and Harbarth et al. [11] showed that starting antibiotic therapy before sample collections may be associated with less-sensitive microorganisms. However, the former only included 76 consecutive patients with ventilator-associated pneumonia [10], and the latter [11] mainly evaluated gram-negative pathogens. Our study expands on these reports, enrolling a large number of patients and analyzing numerous pathogens. To the best of our knowledge, this is the first systematic study evaluating the ecological impact of antimicrobial agents from hospitalized patients 13 based on a large sample size.
From the analysis of the distribution of microorganisms from specimens taken before and after antibiotics therapy, we found that gram-positive organisms had obvious differences between the groups. The results revealed that the percentage of S. aureus was higher in the SAA group, while the isolation rates of CNS and Streptococcus spp. were higher in the SBA group. The differences in S. aureus and Streptococcus spp. could be explained by the fact that Streptococcus spp. antimicrobial susceptibility is significantly higher than that of S. aureus, so the separation rate after treatment may be significantly reduced. However, the reason for the decrease in the separation rate of CNS after the use of antibiotics is unclear. As for gram-negative bacteria, the proportion of non-fermentative bacteria such as Pseudomonas spp., Acinetobacter spp. , and S. maltophilia isolated from the SAA group was higher than that of the SBA group. It is well known that non-fermentative bacteria are opportunistic nosocomial pathogens and are often found in the hospital environment, frequently with multiple-resistance mechanisms. For example, A. baumannii has accumulated resistance to most antibiotics [13][14][15], and in most cases antibiotic treatment should be avoided by virtue of low potential virulence [16]. Additionally, this study also found that the separation rates of H. influenzae and M. catarrhalis were significantly lower after antibiotics administration. It has been speculated that the two bacteria are pathogens of typical community sources, and that they are generally sensitive to most antibacterial agents as is consistent with the present results [17], although recent studies have shown that resistance in these two bacteria is increasing [18]. Our results suggest that insensitive bacteria and common multidrug-resistant (MDR) bacilli in hospitals are more likely to be isolated after antibiotic exposure, which may further result in the use of unnecessary antibiotics and provide selective pressure in favor of MDR organisms.
On the other hand, further subgroup analysis based on specimen types also showed 14 significant differences or different rank orders of microbes derived from respiratory secretions, blood or urine samples taken before or after antibiotic exposure. For example, the isolates of A. baumannii, E. faecium, and E. cloacae in blood samples were different between the two groups, with separation rates increasing after antibiotic therapy. The same phenomenon occurs in urine cultures. The subgroup analysis of urine samples showed that differences between the two groups were significant. Bidell et al. [19] reported that P. aeruginosa isolated from the urine of patients suffering from urinary tract infections is increased in patients with prior antibiotic exposures, but no differences could be found in other pathogens. However, we found that the rates of common microbes such as E. coli, E. faecalis, and P. mirabilis decreased after exposure to antimicrobial agents, whereas the isolation rates of P. aeruginosa and Candida in the SAA group were higher than those isolated from the SBA group. As far as we know, the latter two are not sensitive or affected by commonly used antimicrobial agents. As for respiratory secretions, we found that the proportions of common community acquired pneumonia (CAP) pathogens isolated from respiratory secretions in both groups were less than those reported before [20][21]. Antibiotics having been prescribed before transfer to our hospital, a regional medical center, may account for this difference. Prior antibiotic treatment may reduce respiratory pathogen yield and increase false-positive results [9,[22][23][24][25][26]. In fact, frequent and inappropriate use of antibiotics in primary health care settings in China is a serious problem [27].
However, prior antibiotic exposure seemed have no significant effect on microorganisms isolated from body fluid specimens. Indeed, the few studies investigating the yield of these clinical specimens after antibiotic exposure are inconsistent. In a study with 113 pediatric musculoskeletal infection, Michael A et al [28] showed that antibiotic administration before tissue culture correlated with no detrimental effect on tissue culture 15 sensitivity. Another study assessing the impact of prebiopsy antibiotics on pathogen recovery in hematogenous vertebral osteomyelitis patients also showed that antibiotic exposure before biopsy did not negatively impact pathogen recovery [12]. However, Al-Mayahi et al [29] and Shahi et al [30] both demonstrated the opposite results. Therefore, the role of antibiotic administration in these results is unknown, and further research to help identify possible correlations is needed.
Besides, the study also found the isolation rate of fungi, especially Candida spp., was significantly higher after antibiotic exposure. Usually, Candida spp. can be isolated from airway samples but are not considered to play a significant role in acute illness [31].
Other authors have found that fungal airway colonization is frequent in patients with mechanical ventilation [31][32][33][34]. Therefore, the clinical significance of the higher isolation rate of fungi after antibiotic exposure needs further evaluation.
There were several limitations to this study. First, this was a retrospective study, thus subjecting our results to multiple biases, such as selection bias. Second, despite a large sample, the study was limited only to a single center, which means that the results may not be applicable to other settings. Third, this study was not able to analyze the use of antimicrobial agents before admission and did not analyze whether the antimicrobial agents used for treatment were effective against the cultured microorganisms, and therefore could not speculate on their causal relationship. Finally, this study also did not identify whether these microorganisms were causative pathogens.

Conclusion
we showed that clinical specimens that are easier to collect, such as blood and urine, were more likely to be collected before antibiotic therapy. Further, in most cases, sensitive organisms were more easily isolated before antibiotic use, while drug-insensitive organisms, such as non-fermentative bacteria and fungi, were more frequently isolated after antibiotic exposure. However, this trend was not obvious in body fluid specimens.
Further clinical studies are needed to analyze the effect of antimicrobial agents on changes in antimicrobial susceptibility.
Abbreviations aCDSS: computerized antimicrobial decision-support system; EMRS: electronic medical record system ; LIS: laboratory information system ; PDA :personal digital assistant; SBA: specimen taken before antibiotic therapy; SAA : specimen taken after antibiotic therapy; Declarations Ethics approval and consent to participate  Comparison of microorganisms isolated from aseptic specimens between SBA (top 15) and SAA groups. Abbreviations: SBA, specimen taken before antibiotic therapy; SAA, specimen taken after antibiotic therapy; E. coli, Escherichia coli; K.