Factors Associated With Prolonged Antibiotic Use in the Setting of Suspected Pneumonia and Negative Bronchoalveolar Lavage Cultures

Abstract

Background: Diagnostic criterion for pneumonia includes clinical data and bronchoalveolar lavage cultures (BALCx) to identify pathogens. Although ~60% of BALCx are negative, there may be reluctance to discontinue antibiotics, leading to prolonged antibiotic use (PAU). Objective: The purpose of this study is to compare outcomes of subjects with negative BALCx with PAU versus without prolonged antibiotic use (nPAU). Methods: A retrospective cohort study was conducted including subjects admitted to the intensive care unit (ICU), with suspected pneumonia, and negative BALCx. Data were compared based on length of exposure to antibiotics, PAU (antibiotics >4 days) versus nPAU (antibiotics <4 days). Results: A total of 128 subjects were included, 57 in the PAU group and 71 in the nPAU group. Baseline demographics were similar between groups. Severity of illness measured by multiple organ dysfunction scores at time of bronchoalveolar lavage (BAL) collection to final result showed a statistically significant decrease in the PAU group but not in the nPAU group. No differences were found in ICU days, ventilator-free days, or mortality; however, length of stay was longer for PAU (23 vs. 17, p = .04). In the PAU group, there were fewer BALCx results of “no growth” (23% vs. 45%, p = .04), more positive gram stains (83% vs. 60%, p = .01) and more positive non-BALCx (40% vs. 14%, p = .01). In a multivariate analysis, factors associated with PAU were positive BAL gram stains (adjusted odds ratio [aOR] 3.1, p = .037) and positive non-BALCx (aOR 4.7, p = .002). Conclusion: For subjects with suspected pneumonia and negative BALCx, positive non-BALCx and positive BALCx gram stain influenced the length of exposure of antibiotics.

Introduction

Hospital-acquired infections are a major cause of morbidity, mortality, and require the use of many resources, specifically in intensive care unit (ICU) patients. ¹ Ventilator-associated pneumonia (VAP) occurs in 9% to 40% of intubated patients, accounts for most of antimicrobial use and is a leading cause of death among nosocomial infections.^2,3 This can cause a significant burden on the health care system. In patients with VAP, the duration of ventilation, prolonged antibiotic use (PAU), and length of ICU or hospital stay can increase cost of patient care. ⁴

Ventilator-associated pneumonia is defined as an infection in the lung parenchyma that develops after a patient is intubated and has been on mechanical ventilation for at least 48 hours. ⁵ Patients typically present with at least 2 of the following signs: temperature alteration (<36°C or >38.3°C), a white blood cell count <4000 cells/mm³ or >10 000 cells/mm³, and the existence of purulent sputum. ⁶ It is important to note that the presence of purulent sputum is difficult to identify in patients with VAP due to the patient being mechanically ventilated and therefore unable to cough and clear sputum. When suctioning is available, this can be identified but is often a difficult clinical parameter to obtain in these patients. When patients meet diagnostic criteria for suspected VAP, the recommendation is for culture obtainment and empiric broad-spectrum antibiotic therapy to cover suspected pathogens as delaying antibiotics in critically ill patients has been associated with higher morbidity and mortality.^7,8 Culture obtainment can occur via sputum sample, deep tracheal aspirate or bronchoalveolar lavage (BAL). While the standard use of BAL in all VAP patients is controversial, it is a possible method for pathogen detection in certain patients, especially if intubated. ⁵ If infection is suspected, antibiotics should not be withheld, for a confirmed culture. ⁵ However, de-escalation of antibiotic therapy is recommended following the identification of causative pathogens from final culture results. ⁹

Although there is no defined standard for the diagnosis of VAP, clinical presenting symptoms in addition to a positive culture, such as a BAL, is consistent with a confirmed diagnosis and appropriately identifying VAP.^10,11 In regards to VAP and the use of BAL results, a positive BAL >10 000 CFU/mL is consistent with a confirmed diagnosis, and a negative <10 000 CFU/mL BAL provides rationale to stop antibiotic therapy. ⁵

Clinicians are encouraged to withhold or discontinue antibiotics in the presence of a negative VAP diagnosis, unless there is a risk for additional nonpulmonary infection or clinical need, to decrease prolonged and unnecessary therapy. ⁵ While de-escalation of antimicrobial therapy is an important consideration in ICU care, lack of protocolization as well as other factors such as persistent laboratory or subjective indication of infection, positive gram stain, or recurrent fevers can lead to uncertainty about de-escalation resulting in continued empiric antibiotic therapy even after a negative BAL is received.^12,13 Exposure to extended antibiotic therapy when it is not clearly clinically warranted can lead to unnecessary adverse effects of the therapies themselves, increase the risk of drug interactions, and potentially promote antimicrobial resistance. Prolonged antibiotic use can also increase patient’s risk of developing other superinfections by disrupting the body’s endogenous flora that serves as a protection against pathogens. ⁸ Clostridium difficile often develops in patients on broad-spectrum antibiotic therapy, such as the initial empiric therapy recommended for VAP.^5,14 In addition, no increase in mortality has been demonstrated to result from stopping antibiotic treatment in patients with pneumonia and a negative BAL culture.^15,16 In patients with suspected VAP and a negative definitive culture result, there is concern for continuation of antibiotic therapy and the implications for antimicrobial resistance as well as superinfections.^8,16

The purpose of this study was to evaluate clinical outcomes, specifically ventilator-free days, of subjects prescribed PAU to those not prescribed prolonged antibiotics (nPAU). In addition, we sought to evaluate other clinical outcomes such as length of stay (LOS), mortality, antibiotic utilization, and resource utilization. Finally, we also sought to identify potential factors associated with PAU in subjects suspected to have pneumonia and a negative BAL culture as a hypothesis-generating outcome of interest.

Methods

This institutional review board (IRB)-approved, retrospective cohort study was conducted at a large academic tertiary-care medical center. At the time of the study, the surgical ICUs were separated into a 35-bed surgical ICU and a 14-bed trauma ICU. Subjects were included in the study if they were admitted to either of the surgical ICUs between 2010 and 2012, suspected of having VAP based on an accepted standard of care criteria (VAP was documented in the ICU data repository described below), evaluation by critical care personnel and had a negative BALCx. At the study institution, patient’s underwent BAL culture obtainment when there was clinical suspicion of VAP based on provider discretion and was completed in the ICU. Bronchoalveolar lavages were obtained in the ICU and conducted by a member of the ICU team under the supervision of the attending physician. Bronchoalveolar lavage results were reported by the microbiology lab similar to all other microbiology results, with gram stain, colony count, organism, and sensitivities as the results are available. Negative BALCx was defined by unit specific protocols as 1 of 3 criteria: no growth, ≤10 000 CFU/mL growth, or normal respiratory flora growth. Subjects were excluded if there was documentation of any antibiotic exposure in the 48 hours prior to BAL procedure. This included any antibiotic ordered in the electronic medical record for the patient and included perioperative antibiotic orders that were placed computerized provider order entry system. Subjects were then evaluated based on length of antibiotic exposure following initially suspicion of pneumonia and BAL with PAU defined as greater than 4 days and nPAU defined as less than or equal to 4 days of antibiotics. The definition of nPAU was defined at 4 days due to previous studies identifying 3 or 4 days as defined cutoffs for prolonged antibiotic exposure having an impact on outcomes.^17,18 Our choice of 4 days (over 3 days) was to be more inclusive of situations where final culture results may not allow for clinical decision-making exactly at 72 hours and allows for a more practical application of results.

Data collected included patient demographics including age, gender, pulmonary history, comorbidities, and reason for admission; clinical information including pertinent laboratory data, vital signs, vasopressor exposure, and data used to calculate severity of illness scores at baseline and through hospitalization; culture data including BAL gram stain, colony counts, and non-BALCx positive cultures; clinical outcomes; and antibiotic use data including days of exposure, total antibiotic days for all agents, and cost information. Baseline severity of illness scores included the Trauma Injury and Severity Score (TRISS) ¹⁹ for trauma patients and APACHE II score for all others. Data collection occurred by study investigators using multiple sources including the electronic medical record, the facilities billing database, and a critical care admissions data repository. The Division of Trauma and Surgical Critical Care has created several prospectively collected data repositories. One such repository is an IRB-approved electronic data repository currently maintained for all admissions to either the surgical or trauma ICU and includes significant clinical data and outcomes related to the patient’s ICU stay. Study data were collected and managed using REDCap electronic data capture tools hosted at the study institution. ¹⁷ Multiple organ dysfunction scores (MODS) and clinical pulmonary infections scores (CPIS) were calculated at baseline, at BAL collection and final BALCx result for both groups.

The primary outcome of this study was ventilator-free days between the 2 groups. Secondary outcomes included ICU and hospital LOS as well as change in severity of illness scores from baseline to final BALCx result. Additional outcomes of interest included non-BAL cultures and antibiotic usage data.

Antibiotic usage data are reported in Table 1. Reported data are empiric antibiotic calendar days defined as number of days patient was on empiric antibiotics as well as total antibiotic days which is defined as the total antibiotics used. For example, if a patient was on 2 empiric antibiotic agents then that would be 2 total antibiotic days and 1 antibiotic calendar day. Antibiotic cost was calculated using the Defined Daily Dose (DDD) calculation which is a process developed by the World Health Organization of calculating cost of antibiotic therapy based on a standardized average maintenance dose per day. ²⁰ The online database used to determine standardized DDD is available at https://www.whocc.no/atc_ddd_index/.

Table 1.

Baseline Characteristics.

Baseline Characteristics
	PAU (n = 57)	nPAU (n = 71)
Age (y), (mean ± SD)	55.2 ± 17.4	54.1 ± 18.9
Male gender	70.2%	67.6%
Trauma patients	56.0%	60.5%
Mechanical ventilation	93.0%	93.0%
Pre-existing pulmonary condition	21.5%	22.5%
Chest injury (TICU)	25%	35%
Trauma Injury and Severity Score (TRISS) (TICU), (mean ± SD)	0.564 ± 0.35	0.674 ± 0.31
APACHE II (SICU), (mean ± SD)	19.5 ± 6.7	20.0 ± 5.9
Emergent admission	35.1%	28.2%
BAL Data
	PAU	nPAU
Days to BAL final result, (mean ± SD)	3.3 ± 0.92	3.1 ± 0.93
ICU days to BAL collection, (mean ± SD)	4.2 ± 6.8	4.8 ± 7.2
BAL prior to study inclusion (w/in 72 hours of another)	8.9%	14.1%
Positive gram stain	82.5%	60.0%
Repeat BAL collected (after 72 hours of first)	28.1%	21.1%
Positive repeat BAL	18.8%	33.3%
Other culture-type sent from BAL (i.e. viral and fungal)	8.8%	15.5%
Suspected aspiration event	10.5%	14.3%
Final BAL result
No growth	22.8%	45.1%
<10 000 CFU/mL	61.4%	47.9%
Normal flora	21.1%	22.5%

BAL = bronchoalveolar lavage; PAU = prolonged antibiotic use; SICU = surgical intensive care unit; TICU = trauma intensive care unit.

Finally, a multivariate logistic regression was performed to determine factors associated with prolonged exposure to antibiotics. Variables included in the regression analysis were defined a priori based on other studies looking at impact of PAU or were hypothesized by the authors. Statistical analysis included t-test, Wilcoxon signed-rank test, chi-square test, and logistic regression where appropriate. All results were 2-tailed and statistical significance was defined as a p value ≤ .05. Statistical analysis was completed by the study authors using SPSS, version 21 (IBM Corp. Released 2012. IBM SPSS Statistics for Windows, version 21.0. Armonk, NY: IBM Corp.)

Results

Patient baseline demographics, characteristics, and outcomes are listed in Table 1. Both the PAU (n = 57) and nPAU (n = 71) groups were similar, predominately male gender and had average ages of 55.2 and 54.1 years, respectively. Both groups represented surgically critically ill patients and were similar with regards to pre-existing pulmonary infections, their APACHE II scores, and percent of patients requiring active vasopressors on the date of the BAL final results. The PAU group compared to nPAU, had more days on antibiotics (9 vs. 3 p < .001) and higher antibiotic costs ($775.82 vs. $263.68. p < .001).

The primary and secondary outcomes are reported in Table 2. There was no difference in ventilator-free days, ICU LOS, or hospital mortality. In addition, patients in the PAU had a longer hospital LOS.

Table 2.

Outcomes.

Outcomes	PAU	nPAU	p value
Infectious diseases consultation	7.0%	5.6%	.75
Active vasopressors on date of BAL final result	15.8%	15.5%	.96
Renal replacement therapy	7.1%	2.8%	.25
Days on empiric antibiotics (mean, SD)	9.0 ± 3.4	3.0 ± 0.93	<.001
Total empiric antibiotic days (mean, SD)	17.6 ± 7.1	7.2 ± 3.1	<.001
Ventilator-free days, (mean ± SD)	14.0 ± 9.5	14.7 ± 8.5	.99
ICU LOS, (mean ± SD)	15.6 ± 13.1	11.8 ± 7.8	.31
Hospital LOS, (mean ± SD)	22.6 ± 13.3	17.4 ± 9.1	.04
Hospital mortality	12.3%	12.7%	.95
30-day hospital readmission	11.3%	14.5%	.46

BAL = bronchoalveolar lavage; LOS = length of stay; PAU = prolonged antibiotic use; ICU = intensive care unit.

Baseline severity of illness scores (MODS/CPIS), presented in Table 3, were equivalent for both groups. The reduction in MODS/CPIS scores at BAL collection and final results are presented in Table 3. The mean change in the PAU group scores were statistically significant for both MODS/CPIS scores (−3.0, [p = .02]/−1.4, [p = .02]).

Table 3.

Severity of Illness Scores.

MODS/CPIS Scores
	PAU	nPAU	p value
MODS baseline, median (IQR)	9 (7-12)	9 (5.5-10)	.38
MODS final, median (IQR)	6 (4.75-10.25)	6 (5-9.5)	.68
CPIS baseline, median (IQR)	4 (2-5)	4 (3-5)	.48
CPIS final, median (IQR)	3 (2-4)	3 (1-4)	.56
MODS/CPIS Scores at BAL collection and final result
	Mean difference, (95% CI)	p value
CPIS
PAU	−1.4 (−2.5 to −0.3)	.02
nPAU	0.4 (−0.9 to 1.6)	.52
MODS
PAU	−3.0 (−5.5 to −0.5)	.02
nPAU	−3.3 (−12.7 to 6.1)	.29

BAL = bronchoalveolar lavage; IQR = interquartile range; MODS = multiple organ dysfunction scores; CPIS = clinical pulmonary infections scores; PAU = prolonged antibiotic use; CI = confidence interval.

Positive non-BAL cultures based on site are shown in Figure 1. Prolonged antibiotic use appeared to be associated with more positive non-BALCx sites (40.4% vs. 14.1%. p = .01). Cultures were positive and statistically significant in PAU compared to nPAU, for both blood (22.8% vs. 8.5%, p < .001) and urine sites (17.5% vs. 2.8%. p < .001).

Figure 1.

Distribution of non-BAL culture sources/sites obtained on patients in PAU versus nPAU groups. Logistic regression reported non-BAL positive cultures were associated with PAU. This figure breaks down the different sites these non-BAL cultures were collected from.

Although the PAU group was associated with fewer “No Growth” results (23% vs. 45%, p = .001), the BALCx had more ≤10 000 CFU/mL Enterobacteriaceae growth (21% vs. 8%, p = .04), and more positive gram stains (83% vs. 60%, p = .01). From a multivariate regression model, shown in Table 4, factors associated with PAU were positive BAL gram stains (adjusted odds ratio [aOR] 3.1, 95% confidence interval [CI] = 1.1, 9.2) and positive non-BALCx site (aOR 4.7, 95% CI = 1.8, 12.7); both were statistically significant (p = .037) and (p = .002), respectively.

Table 4.

Results of Multivariate Regression Analysis of Factors Associated With PAU.

Factors Associated with PAU
	Odds ratio (95% CI)
Positive gram stain	3.1 (1.1-9.2)
Other positive cultures	4.7 (1.8-12.7)
“No Growth” on BALCx final result	0.7 (0.3-1.9)
Hospital length of stay	1.0 (0.9-1.0)
≤10 000 CFU/mL—Enterobacteriaceae on BALCx final result	2.2 (0.7-7.1)

BALCx = bronchoalveolar lavage cultures; PAU = prolonged antibiotic use; CI = confidence interval.

Discussion

The results of our study show that in patient with presumed VAP and negative BALCx, there was no difference in ventilator-free days between patients who received more than 4 days of antibiotics versus those who received less than 4 days of antibiotics. Findings also included no difference in ICU LOS, hospital mortality, or 30-day readmission. There was a difference in severity of illness scores in the PAU group from baseline to BAL final result as well as a longer hospital LOS and duration of treatment in the PAU. The severity of illness scores are multifactorial, and it is unclear whether the significant difference between PAU and nPAU is due solely to antibiotic usage. Finally, the findings of a multivariate regression analysis of the data suggest an association between prolonged antibiotics and positive BAL gram stain as well as non-BALCx-positive cultures. The purpose of this study was to evaluate clinical outcomes of patients prescribed PAU versus nPAU in an effort to broaden the body of evidence regarding withholding antibiotics with a negative BAL culture, so that clinicians may comfortably make the best decisions for their patients.

While pneumonia remains an ever-present cause of morbidity and mortality in critically ill patients, refinement of our treatment of these patients must also continue to progress. In patients diagnosed with VAP, the most recent IDSA/ATS guidelines give a strong recommendation for a treatment course of 7 days with agents suitable for adequate empiric coverage of Staphylococcus aureus and Pseudomonas aeruginosa. ⁵ While this study was conducted prior to the publication of these guidelines, evidence at the time suggested a shorter course of antibiotics should be considered with the caveat that some may require longer than 7 days. Patients receiving longer courses of antibiotics are at increased risk of complications, and therefore, every effort should be used to minimize unnecessary exposure to antibiotics while still ensuring effective treatment of definitive infection if present.

Of note, our patient population was mechanically ventilated or had tracheostomy and in an ICU with 24-hour intensivist coverage which made invasive culture obtainment via BAL rapid and streamlined. In situations outside of this scenario, BALCx may be difficult to obtain and more importantly may delay antibiotic initiation. There is controversy about the use of routine BAL procedures for VAP diagnosis due to the potential for delay in antibiotic initiation as well as risk and cost associated with the procedure itself, leading to guideline recommendations for noninvasive culture obtainment for routine diagnosis of VAP. ⁵ If, however, invasive culture obtainment is used, the 2016 IDSA/ATS guidelines recommend to discontinue antibiotics when invasive cultures are negative which indicates that invasive culture results are to be considered definitive. ⁵ This, however, can be a difficult decision for a clinician to make, especially in a complicated and critically ill patient.

Our results suggest that patients in the prolonged exposure group more often had positive BALCx gram stain results even though the final culture results were negative which emphasizes the importance of prioritizing finalized results over preliminary results. This is a meaningful finding because of the frequency in clinical practice of discontinuity between gram stain results and final cultures. This often can lead to difficulty making definitive decisions about antibiotics, especially if patient disposition is unclear or there are subjective concerns about infection. With the increase of rapid culture determination, there is question about the continued utility of gram stain, given its propensity for clouding decision making in certain scenarios.

Our findings suggest that in patients without positive non-BALCx, there was no difference in outcomes or change in severity of illness scores indicating that prolonged antibiotics in these cases were unnecessary. While this is a complicated clinical scenario in many cases, our results assist in supporting the decision to discontinue antibiotics in specific situations where definitive cultures are negative.

When looking at the subset of patients in the PAU group and without a positive non-BALCx (34 of the 57 patients in the PAU group), we found similar outcomes although this analysis would be hypothesis generating and not part of the planned analysis. Ventilator-free days were similar at 15 days, MODS and CPIS scores (median and interquartile range [IQR]) were identical to those found in the full analysis, hospital LOS was 20 days, and ICU LOS was 14 days. In-hospital mortality occurred in 4 of the 34 patients (11.8%) in this subgroup, which is also in line with the 12% seen in the larger group. The similarity of outcomes seen in this subset of patients with prolonged antibiotic exposure and without other positive cultures at the time of BAL obtainment outlines the lack of benefit to continuing antibiotics in the setting of negative cultures.

Our study has a number of notable strengths that makes these data appealing for integration into the body of evidence. First, the patients included in the study are comprised of surgical-intensive and trauma-intensive care patients, who are at a substantially higher risk of infection than the general population. In addition, there is not as substantial a literature base for these specific patient populations compared to other critically ill patient populations. A second strength of this study is the use of severity of illness scores to compare groups. While baseline demographics and clinical features are aspects hopefully within comparable differences in study design, not all studies use multiple indices to stratify the level of illness of their patients. This extra delineation of illness further strengthens the internal validity of the analysis as well as provides a reference to which other providers may compare their critically ill patients. Finally and perhaps most importantly, the patients in this study did not receive antibiotics within the prior 48 hours of BAL, ensuring that any and all cultures retrieved would not be suppressed by any antimicrobial agents.

Despite these strengths, this study has its share of limitations. First, this was a retrospective review and therefore there is risk of bias in clinician decision-making due to lack of ability to capture the reasoning for continuation or discontinuation of antimicrobials. Subjective findings and clinical disposition outside of captured data also make inferences difficult. Due to the select patient population studied, our sample size is relatively modest. Additionally, these results may not be as easily extrapolated to other patient populations beyond surgical intensive and trauma intensive care patients. In addition, our study was not powered to detect secondary and exploratory outcomes and may have been underpowered to detect differences in ventilator days. We did not conduct a sample size calculation, as we were limited to the number of patients available in the database at the time of analysis. Finally, the data collected is a historical sample of patient populations that routinely have changes in standard of practice over the course of many years, and the same results may not be as replicable if standard of care were to drastically change.

The authors also note that the presence of suspected nonpulmonary infections likely influenced some of the clinical decision making, as evidenced by the positive association between non-BALCx-positive cultures and PAU.

As noted above, these conclusions are drawn from a retrospective data set and should be considered hypothesis generating and supportive of previous data reporting likewise results.

This study contributes to the body of literature outlining the importance of limiting antibiotic exposure and identifying causes for prolonged exposure. Limiting antibiotic exposure has implications for LOS, risk of resistance development and cost. Clinical decision making is often multifactorial and involves prioritizing multiple aspects of patient care, including deciphering the relative importance of multiple cultures.

Future research considerations may involve the continued work of definitive rapid culture result reporting as well as more research on outcomes associated with appropriate early discontinuation of antibiotics in patient care. In addition, early detection of non-BALCx-positive cultures would be beneficial for guiding therapy decisions. With risks associated with prolonged exposure to antibiotics, equipping practitioners with evidence-based rationale for definitive de-escalation and discontinuation of antibiotics is essential. Finally, we explored the potential role that severity of illness scores could play as a factor in determining whether de-escalation or discontinuation of antibiotics could occur. While this was not the focus of our study, further evaluation of the use of severity of illness scores in real-time clinical practice would be beneficial for optimizing care.

Conclusion

It is the authors’ conclusion that antibiotics may be safely discontinued in patients with negative BAL cultures in the critically ill surgical or trauma patient. This conclusion directly supports the 2016 IDSA HAP/VAP guideline’s recommendation on invasive cultures. This study demonstrates that for patients with presumed pneumonia and negative BALCx, outcomes were similar between PAU and nPAU groups with the exception of a longer hospital LOS in the PAU group. In addition, positive preliminary BALCx gram stain and positive non-BALCx were associated with prolonged antibiotic exposure. In a subgroup of PAU patients without other positive cultures, there were no changes in outcomes indicating that prolonged antibiotic exposure may not confer additional benefit while exposing the patients to potential harm. Overall, this study underscores the importance of early culture result detection and supports the practice of de-escalation of antibiotics based on culture results and clinical picture.