INVESTIGATIONS ON THE EPIDEMIOLOGICAL DYNAMICS OF THE BACTERIAL COMMUNITIES IN THE HOSPITAL ENVIRONMENT

Background During the second half of the XX Century, a novel type of environment has been created by humans: the big, multi-ward hospital, an infrastructure capable to concentrate into a single building, or a few communicating buildings, thousands of inpatients, interconnected by a dense contact network of persons and a variety of devices and “wares”. This man-made environment crated an environment that offers conditions that could speed the evolution of microorganisms, as a consequence of both the selective pressure deriving form antibiotic treatment (and also other drugs, disinfectants and other chemicals widely used in the hospital) and in relation with the high chance of exchange of genetic material between microorganisms, either within the same species, or among different species. This PhD thesis is part of a long-term project aimed at investigating the hospital as a novel type of environment, that is likely to shape and speed the evolution of bacteria and other microorganisms. In this context, I contributed to the generation of the SkyNet platform (Chapter 2), to retrieve, store and analyse the health data generated every day within the hospital, as well as other information useful for example to reconstruct the patient’s contact networks, and then we tested the platform on hospital acquired infections (HAIs) (Chapter 3, 4, 5). This thesis is thus divided into a total of 6 chapters. Chapter 1 is a brief introduction. Chapter 2 presents the structure and construction of the SkyNet platform. Chapters 3, 4, 5 present specific applications of the platform. A summary of Chapters 2, 3, 4, 5 is as follows. A short conclusion is at Chapter 6. Summary of Chapter 2: SkyNet, an IT platform to monitorHAIs in the hospital environment Background Hospital acquired infections (HAIs) affect about 2,5 million people in Europe each year, with around 90,000 casualties. It is therefore widely recognized that HAIs carry important social and economic repercussions: by considering the associated expenses and gain losses, each HAI case costs around 5,000 euros to the hospital. In view of all these considerations, our aim was to create SkyNet, an innovative genomic epidemiology platform for monitoring HAIs in the hospital environment, designed to help clinical microbiologists in several ways. Materials and methods SkyNet platform is hosted on a server machine. The Relational Database Management System is called SkyNet database (SkyNet DB). The SkyNet Web User Interface (SkyNet WUI) is hosted on the server machine. The SkyNet platform, and the server machine that hosts it, meet all the security requirements necessary to be able to adequately store sensitive data relating to inpatients. Results About 23,452 lines of SQL code has been written from October 2018 to realize SkyNet DB. Since 1st of December 2019, we imported in SkyNet DB from Fondazione IRCSS Policlinico San Matteo of Pavia a total of 48,051,556 records. SkyNet DB consists of: 106 tables, of which 15 are intermediate table. About 33,103 lines of code has been written from October 2018 to realize SkyNet WUI. SkyNet WUI consists in 27 WEB pages. SkyNet WUI has been designed to be used by 4 different types of users: the Administrator user, who has full grants on the WUI; the Advanced user, for affiliated hospitals that have granted full use of the data; the Basic user who is not a user of an affiliated hospital but who can send service requests to the SkyNet platform; the Database Tables Management user, exclusively dedicated to the members of SkyNet group. Conclusions SkyNet represents an innovative data mining based platform on the massive historical and prospective series of information stored in the hospitals’ database, integrated with data from bacterial genomic analyses, in order to discover novel risk factors and predictors of nosocomial infections, and to develop an IT-assisted system to counteract these infections, and antimicrobial resistances. Summary of Chapter 3: Hospital Acquired Infections Risk Surveillance Background Hospital acquired infections (HAIs) represent one of the biggest causes of death in most countries, with the early estimates showing about 2,5 million people affected in Europe every year, and 90,000 of victims. So, the aim of this study was to create a mathematical model able to predict HAIs in Fondazione IRCSS Policlinico San Matteo of Pavia. Materials and methods Cox multivariate regression models with mixed effects was performed by evaluating the proportional hazards assumption, to estimate the risk of experiencing HAIs to Carbapenemsresistant Enterobacterales (CRE), Carbapenems-resistant Pseudomonas aeruginosa (CRPA), Enterobacterales extended-spectrum Beta-lactamases/AmpC Beta-lactamase (ENT ESBLs/AmpC), Methicillin-resistant Staphylococcus aureus (MRSA). Study population was divided according with the main biological sample materials: blood culture (BC), respiratory materials (RM), urine culture (UC), wound swab (WS). Results Data about 49,563 hospitalizations in Fondazione IRCSS Policlinico San Matteo of Pavia from 1st January 2012 to 31st December 2019 were extracted from SkyNet DB. We observed that 2 models provide very good results: the model on RM, relating to the microorganisms CRE showed sensitivity and specificity levels of 82%-76%, respectively, while the model on WS carried out on CRE provided a sensitivity level of 100% and specificity of 82%. We also found 3 models with good performances: 2 models performed on hospitalizations having BC samples in relation to CRE (sensitivity: 60%; specificity: 72%) and CRPA (sensitivity: 67%; specificity: 77%); 1 model performed on hospitalizations having RM samples in relation to CRPA (sensitivity: 68%; specificity: 65%). Conclusions This study showed how the combined use of data mining and statistical techniques of inferential statistics and mathematical models, can lead to the creation of tools that can improve the clinical management of patients in the hospital environment in terms of predicting the HAIs that represent today one of the biggest health problems. Summary of Chapter 4: Hospital Evaluation of Antibiotic Resistance Time Series Background The use of antibiotics is the best tool to defeat bacterial infections. However, the presence of antibiotic resistance (AR), represents the major barrier to the success of an antibacterial therapy. The presence of AR is notoriously associated with an increase in length of stay (LOS) and mortality, with a significant increase in healthcare costs and refunds. Based on all these considerations, the objective of the present study was to evaluate the prevalence of AR time series among isolates from blood culture (BC) samples of hospitalized patients in Fondazione IRCSS Policlinico San Matteo of Pavia. Moreover, we aimed to investigate the impact of AR on LOS and mortality. Materials and methods All Acinetobacter spp., Enterobacterales, Escherichia coli , Klebsiella pneumoniae, Pseudo- monas aeruginosa, Staphylococcus aureus isolates in BC samples from 1st January 2012 till 31st December 2019 were extracted from SkyNet DB. AR to one, two, and more than two drug classes was evaluated per isolate over the years. The evaluation of impact of AR on LOS and mortality was performed by consider the last isolation for each hospitalization. For each analysis, the average LOS (ALOS) was calculated. Hospitalizations were then grouped into <= ALOS and > ALOS according to their LOS. Results We analysed 2,220 BC isolations, performed in 1,959 hospitalizations, referring to 1,880 patients. Concerning the AR to >2 drug classes, we observed a significant decrease in the percentage of isolates for Acinetobacter spp. (Figure 4.1(a)), Enterobacterales (Figure 4.1(b)), Pseudo- monas aeruginosa (Figure 4.1(e)), Staphylococcus aureus (Figure 4.1(f)). Regarding the impact of AR on LOS, we found that the prevalence of hospitalizations having a LOS higher than ALOS, significantly increased by increasing the number of AR drug classes for Acinetobacter spp. (Figure 4.8(a)), Escherichia coli (Figure 4.8(c)), Klebsi- ella pneumoniae (Figure 4.8(d)), Pseudomonas aeruginosa (Figure 4.8(e)), Staphylococcus aureus (Figure 4.8(f)). About the impact of AR on mortality, we found that the prevalence of hospitalizations in which the patient died, significantly increased by increasing the number of AR drug classes for Acinetobacter spp. (Figure 4.9(a)), Klebsiella pneu- moniae (Figure 4.9(d)), Pseudomonas aeruginosa (Figure 4.9(e)), Staphylococcus aureus (Figure 4.9(f)). Conclusions Our study showed a general significant decrease over the years of AR in BCs, with an increase of isolates without resistance. The significant decrease in resistance mainly regarded the isolates with more than two resistance classes. However, the few cases of isolates with multi-drug resistant should not be underestimated as our data showed that the increased resistance is significantly associated with an increase in LOS, but above all, with a higher mortality rate. Summary of Chapter 5: VAP due to MRSA vs. MSSA: what should guide empiric therapy in ICU environment? Background The guidelines on ventilator-associated pneumonia (VAP) recommend an empiric antibiotic therapy against Methicillin-resistant Staphylococcus aureus (MRSA) rather than Methicillin-susceptible Staphylococcus aureus (MSSA) if more than 20% of Staphylococcus aureus isolates associated with VAP in the unit are MRSA. However, the real need of this choice has been poorly verified. Firstly, this study evaluated MRSA and MSSA VAP prevalence over the years from 2012 till 2021 in Fondazione IRCSS Policlinico San Matteo of Pavia. Secondly, we want to compare patients with MRSA VAP to those with MSSA VAP in terms of length of stay (LOS) and in-hospital mortality. Finally, we wanted to assess the clinical value of the MRSA nasal swab screening, in in either predicting or conversely ruling out MRSA VAP. Materials and methods Data of positive bronchoalveolar lavage (BAL) samples of patients admitted to the Intensive Care Unit (ICU) from 1st January 2012 to 31st December 2021 were retrospectively extracted from the SkyNet DB. According to the new National Healthcare Safety Network definition, only the BAL from patients with imaging test results, signs and symptoms consistent with pneumonia were included. We excluded all the tube-colonizations. The trend of positive BAL for MRSA or MSSA over the years and the differences in prevalence were evaluated by Chi-square tests for trend. The impact on LOS and in-hospital mortality was evalutated by Chi-squared test or Fisher’s Exact test as appropriate. Hospitalizations were stratified as follows: LOS <= the average LOS (ALOS) and LOS > ALOS. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the MRSA nasal swab were calculated. 6 Abstract of the thesis Results Overall, data about 1,461 positive BAL samples were extracted (Table 5.1). Among 1,461 positive BAL, 170 (11.6%) were positive for MRSA or MSSA (Table 5.2). Among VAP due to Staphylococci (N=170), the prevalence of MSSA significantly increased over the years from 56.5% in 2012 to 85% in 2021, by contrast the prevalence of MRSA significantly decreased from 43.5% in 2012 to 15% in 2021 (p=0.038; Figure 5.1). Moreover, there was a general downward trend in MRSA prevalence (from 9.4% in 2012 to 1.3% in 2021, p=0.001; Figure 5.2(a)), while MSSA remained fairly steady over time (from 12.3% in 2021 to 7.1% in 2021, p=0.218; Figure 5.2(b)). Having a VAP due to MRSA did not have any impact on LOS and mortality. The MRSA nasal swab testing demonstrated a 42.1% sensibility and 98.4% specificity, with a PPV of 36.4% and a NPV of 98.7% (Table 5.4). Conclusions Our results showed that, despite a downward trend in prevalence of VAP due to MRSA over the last 9 years, it has overall remained above 20% in the ICU environment of Fondazione IRCSS Policlinico San Matteo of Pavia. We want to highlight that MRSA nasal colonization, which is a recognised risk factor for MRSA VAP, has a significantly high NPV in our analysis. This finding brings compelling thoughts in terms of antimicrobial stewardship, as a negative MRSA nasal swab may be used to rule out MRSA VAP, and consequently guide clinicians’ decisions on empirical treatment.