Ventilator-Associated Pneumonia, Multidrug-Resistant Bacteremia and Infection Control Interventions in an Intensive Care Unit: Analysis of Six-Year Time-Series Data

Abstract: Background Ventilator-associated pneumonia (VAP) occurs more than 48h after mechanical ventilation and is associated with a high mortality rate. The current hospital-based study aims to investigate the association between VAP rate, incidence of bacteremia from multidrug-resistant (MDR) pathogens, and infection control interventions in a single case mix ICU from 2013 to 2018. Methods: The following monthly indices were analyzed: (1) VAP rate; (2) use of hand hygiene disinfectants; (3) isolation rate of patients with MDR bacteria; and (4) incidence of bacteremia/1000 patient-days (total cases, total carbapenem-resistant cases, and carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa, and Klebsiella pneumoniae cases separately). Results: Time trends of infection control interventions showed increased rates in isolation of patients with MDR pathogens (p <0.001) and consumption of hand disinfectant solutions (p =0.001). The last four years of the study an annual decrease of VAP rate by 35.12% (95% CI: −53.52 to −9.41; p =0.01) was recorded, which significantly correlated not only with reduced trauma and cardiothoracic surgery patients (IRR:2.49; 95% CI: 2.09–2.96; p <0.001), but also with increased isolation rate of patients with MDR pathogens (IRR: 0.52; 95% CI: 0.27–0.99; p = 0.048), and hand disinfectants use (IRR: 0.40; 95% CI: 0.18–0.89; p =0.024). Conclusions: Infection control interventions significantly contributed to the decrease of VAP rate. Constant infection control stewardship has a stable time-effect and guides evidence-based decisions.


Introduction
Hospital-acquired pneumonia is an infection of the pulmonary parenchyma caused by pathogens that are present in hospital settings with an incubation period of at least 2 days. Ventilator-associated pneumonia (VAP) develops in intensive care unit (ICU) patients who have been mechanically ventilated for at least 48h [1]. Patients receiving mechanical ventilation are at high risk for complications, disability and poor outcome, longer stays in the ICU and hospital, and increased healthcare costs [2].
The accurate estimation of the incidence of VAP remains a challenge mainly because of subjective interpretation of clinical findings as well as because the diagnosis is made based purely on clinical criteria or via invasive lower airway cultures [3]. The type of ICU and patients also may impact the incidence of VAP; i.e., trauma and cardiothoracic surgery patients are at exceedingly high risk for VAP [4]. The United States National Healthcare Safety Network surveillance for the year 2012 estimated that the incidence of VAP ranged from 0.0-4.4 per 1000 ventilator-days depending on hospital unit [5]. According to the 2012-2017 International Nosocomial Infection Control Consortium surveillance study conducted in 523 ICUs in 45 countries worldwide, the mean incidence of VAP was estimated at 14.1 per 1000 ventilator-days [6]. However, as high as 18.3 episodes of VAP per 1000 ventilator-days were reported in a prospective observational study conducted in 27 ICUs in nine European countries in 2017 [7].
The microbial agents of VAP vary considerably and information about the microbiology of VAP serves to guide optimal antibiotic therapy. Most cases of VAP are caused by bacterial pathogens that normally colonize the oropharynx and gut, or are transmitted through healthcare personnel from other patients or environmental surfaces [8]. Antibioticresistant pathogens such as Pseudomonas and Acinetobacter species and methicillin-resistant Staphylococcus aureus are much more common after prior antibiotic treatment or prolonged hospitalization or mechanical ventilation [4,8].
The association between healthcare-associated infections and infection control interventions was investigated recently using observational and interrupted time-series analyses [9,10]. Various preventive healthcare strategies have been particularly assessed for VAP using this method [11,12]. The aim of the current study was to evaluate the association between the incidence of VAP, the incidence of MDR bacteremia, and specific infection control interventions in an adult ICU in a tertiary-care hospital in Greece.

Setting
This is a hospital-based study conducted prospectively from January 2013 to December 2018 in a 20-bed adults' ICU of a 300-bed private tertiary-care hospital in Athens, Greece.

Interventions
Interventions were implemented throughout the study period and evaluated every month. Interventions consisted of: (1) surveillance of carbapenem-resistant Acinetobacter baumannii, Pseudomonas aeruginosa, and Klebsiella pneumoniae; (2) implementation of a VAP bundle, which included head elevation to 30 • -45 • , daily sedation vacation, daily assessment of readiness to wean, peptic ulcer disease prophylaxis, deep venous thrombosis prophylaxis, endotracheal tube aspiration techniques to keep ventilator circuits clean and dry, mouth hygiene through chlorhexidine 2%, and hand washing before and after providing healthcare to patients; (3) promotion of hand hygiene for all ICU patients; and(4) multi-body-site colonization screening cultures (pharyngeal, axillary-rectal, nasal) and isolation of patients with MDR bacteria.

Data Collection
Data were collected prospectively and monthly during the study period: use of hand disinfectant solutions from Pharmacy Department; number of VAP cases from Nurse Clinical Infectious Diseases Department; number of bacteremia from Microbiology Department. During the study period no changes were made in medical/nurse hospital quality procedures and laboratory diagnostic procedures.

Diagnostic Criteria for VAP
VAP was diagnosed based on a modified combination of both the 2013 Center of Disease Control and Prevention and Johanson criteria. A patient would be diagnosed with VAP if chest X-ray showed new infiltrates and at least two of the following six criteria were present: (1) fever (temperature > 38 • C); (2) abnormal white blood cells (WBCs) count; whether leukocytosis (>12,000/µL) or leucopenia (<2000/µL); (3) altered respiratory secretions; (4) elevated C-reactive protein (CRP) > 3 mg/L; (5) positive blood or respiratory culture; and (6)

Detection of Bacteremia and Microbial Resistance
Bacteremia was detected through Gram stains and blood cultures. The isolation, identification, and antibiotic susceptibility testing were completedusing the automated VITEK 2 system (Biomerieux, Marcy-l'Étoile, France) and the resistance of selected bacteria to specific antibiotics was set out by using CLSI breakpoints. The assayused to determine the susceptibility of bacteria was recorded (Kirby-Bauer test, MIC semi-automated testing, E-test).

Definitions
VAP rate per month was defined as the number of VAP episodes per 1000 ventilatordays. Isolation rate of patients with MDR bacteria was defined as percentage of isolated patients with MDR per admissions (% isolations/admissions). Bacteremia was defined as a laboratory-confirmed bloodstream infection, either primary (not related to an infection at another body site) or secondary (thought to be seeded from a site-specific infection at another body site) [14]. New episode of bacteremia, within a month period, was defined due to a different pathogen strain or due to the same pathogen strain but with different phenotype of resistance. The incidence of total bacteremia is the sum of total Gram-positive and Gram-negative bacteremia. The incidence of total carbapenem-resistant Gram-negative bacteremia is the sum of the incidence of carbapenem-resistant A. baumannii, P. aeruginosa, and K. pneumoniae bacteremia. The incidence of total resistant Gram-positive bacteremia is the sum of the incidence of methicillin-resistant S. aureus plus vancomycin-resistant Enterococcus bacteremia. Hand hygiene includes the use of scrub disinfectant solutions with chlorhexidine, alcohol 70% disinfectant solutions with chlorhexidine, and/or simple soap.

Statistical Analysis
Investigation of time trends in the intervention and outcome variables during study period was initially performed, constituting the dependent variables. Time since the beginning of the study (in months) was the independent variable in the regression models and was entered through appropriate restricted cubic splines. Fourier series terms of time (1st and 2nd order) were also entered in the models to capture potential seasonality effects. In all cases, standard errors (SE) and corresponding 95% confidence intervals (CI) were derived using the robust (sandwich) variance estimator to adjust potential violations of models' assumptions. Estimated values for start and end of the study period and corresponding 95% CIs were estimated through a simplification of the models where spline time terms were replaced by a single linear time trend or two piecewise linear terms to capture average long-term trend. A linear regression model was used for disinfectants consumption. For cases where the outcome of interest was VAP or bacteremia rates, Poisson regression models were used with number of cases as dependent variable and appropriate number of ventilation-days or patient-days, respectively, used as an offset after logarithmic transformation. For cases in which the percentage over total number of hospitalizations was the outcome of interest (isolations), binomial regression models were used with number of cases as dependent variable and the appropriate number of hospitalizations as binomial denominator. Associations between outcomes and interventions were investigated by introducing appropriate independent variables into the models. The effects of the independent variables were initially tested separately for current (month 0) and lagged values (months -1, -2 and -3). If the effects were statistically significant (p < 0.05) or indicative (0.05 < p < 0.10) for more than one case (e.g., in month 0 and in month -1) and association direction was the same (e.g., positive for both), average value was used as independent variable. In cases where direction of the association was different (e.g., positive for month 0 and negative for month -1), results of the respective models are presented separately. All p-values reported throughout the manuscript have not been adjusted for multiple testing. Analyses were conducted using Stata version 14.2 (Stata Corp., College Station, TX, USA).

Distribution of admissions per type of ICU patients is shown in
From the annual number of admissionsin 2013, cardiovascular surgery and trauma patients accounted for 75.36% of ICU admissions and incidence VAP rate was estimated at 7.08 cases per 1000 ventilator-days (Tables 1 and 2 (Table 2). Table 3 shows the six-year time trends of the implemented interventions in the ICU. There was a statistically significant increase in the percentage of isolation of patients with MDR pathogens per 100 admissions(p < 0.001). Regarding hand hygiene, the use of scrub disinfectant solutions increased significantly (p = 0.001), while no significant difference in the use of alcohol disinfectant solutions (p = 0.286) was observed. Overall, the use of all hand disinfectant solutions in the ICU increased significantly (p = 0.001). Table 4 shows six-year time trends of the incidence rate of VAP and bacteremia in the ICU. There was a significant decrease of the incidence of VAP rate by 35.12% per year (95% C.I.: −53.52% to −9.41%; p = 0.01). Overall, there was a significant increase in the incidence of total ICU bacteremia (p < 0.001), but no significant change in the incidence of total carbapenem-resistant Gram-negative pathogens and separately. From the relative change per year, there was a significant increase in the incidence of total carbapenem-resistant A. baumannii bacteremia only the last year of the study (p = 0.003), and of total carbapenemresistant K. pneumoniae bacteremia the first two years of the study (p-value = 0.030). For resistant Gram-positive pathogens, the incidence was very low, almost zero for most time of the study, so a linear model could not be applied. Table 5 shows the correlation between incidence of VAP, incidence of different bacteremia and type of ICU patients. For carbapenem-resistant A. baumannii bacteremia, every increase in their incidence per 1/1000 patient-days, current month, resulted in a statistically significant increase per 17% in the incidence of VAP rate (p = 0.045). For carbapenemresistant K. pneumoniae bacteremia, every increase in their incidence per 1/1000 patient-days, three months earlier, resulted indicatively in 24% increase in VAP rate (p-value = 0.086). For total cardiothoracic surgery and trauma patients every increase per number of 10 resulted in 149% increase in VAP rate (p < 0.001). The same effect was observed for cardiothoracic surgery patients (p < 0.001) and trauma patients separately (p = 0.002). Table 6 shows the correlation of VAP rate with infection control interventions. For every increase per 10% of isolations of patients with MDR pathogens, two months earlier, resulted in a significant decrease per 48% in VAP rate (p =0.048). Regarding the use of hand disinfectant solutions, the correlation was also negative for current month. Every increase in the use of scrub disinfectant solutions and all hand disinfectant solutions per 10 L/1000 patient-days resulted in a statistically significant decrease in VAP rate per 15% (p =0.028) and 60% (p =0.024), respectively. However, there was no significant association with the use of alcohol disinfectant solutions (p =0.079).

Discussion
This is a six-year observational study aiming to assess the relationship between infection control interventions and the incidence of VAP in a 20-bed adults ICU in Greece. From 2013 to 2018 a VAP-bundle was implemented in the ICU continuously audited to promote prevention measures along with an infection control program to promote hand hygiene, contact precautions, and strict isolation guidelines for patient with MDR pathogens.In 2013 and 2014 up to 80% of ICU admissions were trauma and cardiothoracic surgery patients, with an annual VAP rate of 7.08 and 7.377 cases per 1000 ventilator-days, respectively. The incidence of VAP changed dramatically downwards the following years to 21.29% in 2018, with annual VAP rate dropping to 1.418 cases per 1000 ventilator-days. Published data indicate that these groups of ICU patients demonstrate high risk for VAP which may contribute to excess mortality, even though these patients are often otherwise healthy [4].
From a multicenter study in Western European cardiothoracic ICUs, VAP was the most frequent nosocomial infection occurred in 2.1% of patients for an overall estimated rate of 13.9 VAP cases per 1000 ventilator-days [15]. Most cases of VAP are caused by bacterial pathogens that normally colonize the oropharynx and gut [8]. In the majority of patients with the proper functioning immunity, this microbial virulence alone does not account for the virulence of a disease [16,17]. The largestpart of the virulence of an infectious or chronic non-infectious disease is the result of the overactive inflammation response of the immune system, which turns local transient inflammation into systemic or chronic inflammation [18,19]. Overnutrition might be the driving force for such inflammatory transition, and slightly undernutrition may be preferred in reducing hyperinflammation [20,21].
Apart from the underlying condition for which the patient requires ventilation, an additional risk factor for VAP is the increased mechanical ventilation time and prolonged length of hospital stay [22]. In our study the duration of a patient's mechanical ventilation is not consistent with VAP rate, with the mean ventilator-days increasing until 2016 and decreasing afterwards until 2018. In particular, 2016 had the highest mean value for the duration of mechanical ventilation and the lowest VAP rate, indicating indirectly the impact of preventive measures and infection control interventions along with the change of ICU type of patients.
Early clinical practice guidelinesfor VAP were based primarily on experts' opinion derived from clinical experience, but throughout the years, the infectious diseases societies joined the respiratory, thoracic, and intensive care medicine associations to release shared documents [23,24]. In addition, novel approaches have been investigated for the reduction of VAP in the hospital setting using interrupted time-series method [19,20], such aschlorhexidine gluconate bathing [25] and dual hand hygiene audit [26]. However, VAP remains a frequent hospital-acquired infection worldwide and the increasing prevalence of MDR bacteria further drives the need to explore advances in diagnostic and empiric treatment strategies [27,28]. National surveillance data in Greece show a high prevalence of carbapenem-resistant pathogens in acute-care hospitals in this country, constituting a significant public health problem [29]. Additionally, in Greece the estimated incidence of HAIs accounted for 4.3% of cases, and in Europe for 3.7% of cases, according to two European point prevalence surveys conducted in acute care hospitals in 2016 to 2017 [30].
From our study, the most significant interventions in the ICU were the increased isolation rate of patients with MDR pathogens and the increased consumption of all hand disinfectant solutions, indicating adherence to the infection stewardship program. Furthermore, the most statistically significant outcome was the annual significant decrease of 35.12% in the incidence of VAP. To the best of our knowledge, our study provides the first-time data about the correlation of VAP not only with different types of ICU patients but also with different bacteremia and overall infection control interventions in an ICU. Correlation results of the applied time-series analysis showed that the reduction of trauma and cardiovascular surgery patients admitted in the ICU, the isolations of patients with MDR pathogens, and hand hygiene adherence guidelines resulted in a continuous decrease in VAP rate for four consecutive years.
As far the type of bacteremia is concerned, for carbapenem-resistant A. baumannii bacteremia, every increase in their incidence the current month alsoresulted in a significant increase in VAP rate, indicating that contact precautions would be beneficial on top of isolation measures, particularly for this pathogen. Recent publication using time series analysis did not show any difference regarding incidence of methicillin-resistant S. aureus or vancomycin-resistant Enterococcus bacteremia after cessation of contact precautions [31,32]. In our ICU, the increase of the incidence of carbapenem-resistant Gram-negative pathogens was not significant while the incidence of resistant Gram-positive pathogens was almost zero for most time of the study. In order to further reduce the incidence of healthcareassociated infections and particularly the prevalence of carbapenem-resistant pathogens, it is important to guide evidence-based decisions and implement tailored infection control policies [29,30].Our study has several strengths. We applied time series analysis to investigate the potential correlation between the incidence of VAP, infection control interventions, and MDR bacteremia. A clear strength is the six-year time period and the prospective collection of data. A potential limitation is the mixed type of patients admitted in our ICU, including cardiothoracic surgery and trauma patients. Seeing that the issue of adjusting p-values for multiple testing is controversial, we chose to report unadjusted p-values, thus some inflation of the Type I error beyond the typical 0.05 level cannot be excluded [33].

Conclusions
During a six-year study period we investigated the effect of infection control interventions and outcomes in VAP rate using time series data. Primarily, the reduction of trauma and cardiovascular surgery patients admitted in the ICU and secondarily the isolations and hand hygiene adherence guidelines resulted in a continuous decrease in VAP rate for four consecutive years. Favorable outcomes of infection control interventions implemented constantly in the ICU, such asthe increased isolation rate of patients with MDR pathogens and the increased consumption of all hand disinfectant solutions, harbored breakthrough infections from resistant pathogens. Further contact precaution measures could be applied for Gram-negative resistant pathogens in order to ameliorate the robust and constant effect of the infection stewardship program. Time series analysis is an important and safe tool to measure and evaluate data and interventions over time and to guide evidence-based decisions and infection control policies. .Since this study anddata collection was based entirely on data abstraction from existing medical or laboratory record, with no interaction with the human subjects concerned and with no collection of identifiable private information, an informed consent is not required from the patients.

Data Availability Statement:
The data that support the findings of this study are available on request from the corresponding author, A.T. The data are not publicly available as data disclosure requires permission and ethical approval from Medical Ethical Committee of Athens Medical Center, Athens, Greece.