Bayesian Networks in the Management of Hospital Admissions: A Comparison between Explainable AI and Black Box AI during the Pandemic.

Artificial Intelligence (AI) and Machine Learning (ML) approaches that could learn from large data sources have been identified as useful tools to support clinicians in their decisional process; AI and ML implementations have had a rapid acceleration during the recent COVID-19 pandemic. However, many ML classifiers are "black box" to the final user, since their underlying reasoning process is often obscure. Additionally, the performance of such models suffers from poor generalization ability in the presence of dataset shifts. Here, we present a comparison between an explainable-by-design ("white box") model (Bayesian Network (BN)) versus a black box model (Random Forest), both studied with the aim of supporting clinicians of Policlinico San Matteo University Hospital in Pavia (Italy) during the triage of COVID-19 patients. Our aim is to evaluate whether the BN predictive performances are comparable with those of a widely used but less explainable ML model such as Random Forest and to test the generalization ability of the ML models across different waves of the pandemic.

Prognostic Significance of NLR and PLR in COVID-19: A Multi-Cohort Validation Study.

INTRODUCTION: Recent studies have highlighted the prognostic value of easily accessible inflammatory markers, neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR) for predicting severe outcomes in patients affected by Coronavirus disease 2019 (COVID-19). Our study validates NLR and PLR cut-off values from a prior cohort at IRCCS Policlinico San Matteo (OSM) of Pavia, Italy, across two new cohorts from different hospitals. This aims to enhance the generalizability of these prognostic indicators. METHODS: In this retrospective cohort study, conducted at Milan's Ospedale Luigi Sacco (OLS) and IRCCS Ospedale Maggiore Policlinico (OMP) hospitals, we assess the predictive capacity of NLR and PLR for three main outcomes-non-invasive ventilation (NIV) or continuous positive airway pressure (CPAP) usage, invasive ventilation (IV), and death-in patients with COVID-19 at admission. For each outcome, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were computed separately for male and female cohorts. Distinct NLR and PLR cut-off values were used for men (7.00, 7.29, 7.00 for NLR; 239.22, 248.00, 250.39 for PLR) and women (6.36, 7.00, 6.28 for NLR; 233.00, 246.45, 241.54 for PLR), retrieved from the first cohort at OSM. RESULTS: A total of 3599 patients were included in our study, 1842 from OLS and 1757 from OMP. OLS and OMP sensitivity values for both NLR and PLR (NLR: 24-67%, PLR: 40-64%) were inferior to specificity values (NLR: 64-76%, PLR: 55-72%). Additionally, PPVs generally remained lower (< 63%), while NPVs consistently surpassed 68% for PLR and 72% for NLR. Finally, both PLR and NLR exhibited consistently higher NPVs for more severe outcomes (> 82%) compared to NPVs for CPAP/NIV. CONCLUSIONS: Consistent findings across diverse patient populations validate the reliability and applicability of NLR and PLR cut-off values. High NPVs emphasize their role in identifying individuals less likely to experience severe outcomes. These markers not only aid in risk stratification but also guide resource allocation in emergencies or limited-resource situations.

Daptomycin Exposure Prediction With a Limited Sampling Strategy.

BACKGROUND: Daptomycin is a cyclic lipopeptide antibiotic used to treat serious infectious endocarditis caused by Staphylococcus aureus . The pharmacodynamic parameter correlating best with efficacy is the ratio of the estimated area under the concentration (AUC 0-24 )-time curve to the minimum inhibitory concentration. The aim of the study is to develop a limited sampling strategy to estimate AUC 0-24 using a reduced number of samples. METHODS: Sixty-eight daptomycin AUC 0-24 values were calculated for 50 White patients who underwent treatment for at least 5 consecutive days. Plasma concentrations were detected using a validated high-performance liquid chromatography-tandem mass spectrometry analytical method, with daptomycin-d5 as an internal standard. Multiple regression was used to evaluate the ability of 2 concentration-time points to predict the AUC 0-24 calculated from the entire pharmacokinetic profile. Prediction bias was calculated as the mean prediction error, whereas prediction precision was estimated as the mean absolute prediction error. The development and validation datasets comprised 40 and 10 randomly selected patients, respectively. RESULTS: The AUC 0-24 (mg*h/L) was best estimated using the daptomycin trough concentration and plasma concentrations detected 2 hours after dosing. We calculated a mean prediction error of 1.6 (95% confidence interval, -10.7 to 10.9) and a mean absolute prediction error of 11.8 (95% confidence interval, 5.3-18.3), with 73% of prediction errors within ±15%. CONCLUSIONS: An equation was developed to estimate daptomycin exposure (AUC 0-24 ), offering clinical applicability and utility in generating personalized dosing regimens, especially for individuals at high risk of treatment failure or delayed response.

Optimizing Antibiotic Therapy for Intravenous Drug Users: A Narrative Review Unraveling Pharmacokinetics/Pharmacodynamics Challenges.

Intravenous drug users (IVDUs) face heightened susceptibility to life-threatening gram-positive bacterial infections, particularly methicillin-resistant Staphylococcus aureus (MRSA). While the standard antibiotic dosing strategies for special patients, such as obese or critically ill individuals, are known to be inadequate, raising concerns about treatment efficacy, a similar sort of understanding has not been assessed for IVDUs yet. With this in mind, this review examines the pharmacokinetic/pharmacodynamic characteristics of antibiotics commonly used against gram-positive bacteria in IVDUs. Focusing on daptomycin, vancomycin, teicoplanin, aminoglycosides, and the novel lipoglycopeptide dalbavancin, the study reveals significant pharmacokinetic variations in IVDUs, suggesting the need for personalized dosing. Concomitant opioid substitution therapy and other factors, such as malnutrition, contribute to altered pharmacokinetics/pharmacodynamics, emphasizing the importance of targeted therapeutic drug monitoring. Overall, our study calls for increased awareness among clinicians regarding the unique pharmacokinetic/pharmacodynamic challenges in IVDUs and advocates for tailored antibiotic dosing strategies to enhance treatment outcomes in this marginalized population.

The Prognostic Role of Diagnostic Criteria for COVID-19-Associated Pulmonary Aspergillosis: A Cross-Sectional Retrospective Study.

Several criteria exist to diagnose pulmonary aspergillosis with varying degrees of certainty in specific populations, including oncohaematological patients (EORTC/MSG), ICU patients (mAspICU) and COVID-19 patients (ECMM). At the beginning of the pandemic, however, the diagnosis of COVID-19-Associated Pulmonary Aspergillosis (CAPA) could not be performed easily, and the decision to treat (DTT) was empirical. In this cross-sectional retrospective study including patients with SARS-CoV-2 infection and suspicion of CAPA, we studied the concordance between the DTT and the three diagnostic criteria using Cohen's coefficient, and then we identified the factors associated with the DTT and corrected them by treatment to study the influence of the diagnostic criteria on survival. We showed good concordance of the DTT and mAspICU and ECMM criteria, with "compatible signs", "positive culture" and "positive galactomannan" influencing the DTT. Treatment also showed a positive effect on survival once corrected for a putative, possible or probable diagnosis of CAPA using mAspICU and ECMM criteria. We conclude that EORTC/MSGERC are not considered applicable in clinical practice due to the lack of inclusion of signs and symptoms and do not lead to improved survival. mAspICU and ECMM criteria showed a good degree of agreement with the DTT and a positive correlation with patient recovery.

Impact of proton pump inhibitors on the onset of gastrointestinal immune-related adverse events during immunotherapy.

INTRODUCTION: The gut microbiota (GM) can influence the pathogenesis of immune-mediated adverse events (irAEs). Proton pump inhibitors (PPIs) can affect the integrity of GM, but their role in promoting irAEs is still poorly understood. METHODS: In this retrospective single-center cohort study, the primary endpoint was the evaluation of the incidence of gastrointestinal (GI) irAEs in cancer patients on PPIs (exposed) versus cancer patients who were not on PPIs (unexposed). RESULTS: Three hundred and sixty three patients' records (248 M/115F, median age 69) were reviewed. Twenty-three exposed patients (92%) developed GI irAEs while only two unexposed patients (8%) developed GI irAEs (hazard ratio [HR] 13.22, 95% confidence interval [CI] 3.11-56.10, p < 0.000). This HR was confirmed after weighting for the propensity score (HR15.13 95% CI 3.22-71.03, p < 0.000). CONCLUSION: Chronic PPI use is associated with an increased risk of GI irAES.

Fungal Osteomyelitis: A Systematic Review of Reported Cases.

Fungal osteomyelitis is considered a rare disease, and the published literature mainly comprises case reports, case series and narrative reviews. A systematic review was undertaken to provide a practice-based global perspective on this disease, focusing on epidemiology and treatment strategies. We searched MEDLINE, EMBASE and Cochrane Library between the 3rd and 8th of March 2023 using a predefined search string. We included studies with at least one patient with a diagnosis of fungal osteomyelitis published before the 1st of January 2023. We included all study designs except for reviews, and we excluded non-English languages and grey literature. After exclusion, 678 studies, mostly case reports, were included. Descriptive analysis was performed on 1072 patients. The most common aetiological agent was Aspergillus (26.5%), followed by Candida (20.7%) and Mucor (16.8%), and the bones most frequently involved were the vertebrae. We described the characteristics of patients divided by site of infection, and we found that diabetes mellitus, disseminated fungal infection, surgery and local lesion were major risk factors. We also successfully associated duration of treatment with outcome. We provided a general overview of this rare disease, and we highlighted the need for high-quality investigations on the subject.

Gut Microbiota and B Cell Receptor (BCR) Inhibitors for the Treatment of Chronic Lymphocytic Leukemia: Is Biodiversity Correlated with Clinical Response or Immune-Related Adverse Event Occurrence? A Cross-Sectional Study.

Several studies have strengthened the link between the gut microbiota (GM) and the response to immunotherapy in patients with tumors, highlighting the potential role of GM as a biomarker of response. Targeted therapies including B-cell receptor (BCR) inhibitors (BCRi) represent the newest approach to the treatment of chronic lymphocytic leukemia (CLL); however, not all patients achieve a satisfactory response, and immune-related adverse events (irAEs) can also impact the efficacy. The aim of the study was to compare GM biodiversity in patients with CLL, treated with BCRi for at least 12 months. Twelve patients were enrolled: 10 patients in the responder group (R) and 2 patients in the non-responder group (NR). We identified seven patients (58.3%) who experienced adverse reactions (AE). Although we did not observe a significant difference across the study population in terms of relative abundance and alpha and beta diversity, we found a differing distribution of bacterial taxa between the analyzed groups. We noted a higher level of the class Bacteroidia and the order Bacteroidales in the R group, and an inversion in the Firmicutes and Bacteroidetes ratio in the AE group. No prior studies have focused on linking GM and response to BCRi in these patients. Although the analyses are preliminary, they provide suggestions to guide future research.

The possible role of serum bactericidal titres in long-term suppressive antibiotic treatment for infective endocarditis: report of three cases.

INTRODUCTION: long term suppressive antibiotic treatment may be the only feasible option for patients with infective endocarditis (IE) not suitable for surgery. CASE REPORTS: we present three cases of prosthetic valve endocarditis (PVE) caused by Enterococcus faecalis and Streptococcus gallolyticus which could not undergo surgical intervention due to high risk. Despite this, patients were successfully managed only by medical approach. Following intravenous targeted antimicrobial therapy, patients received chronic suppressive antimicrobial therapy (SAT) for at least twelve months with oral amoxicillin. In all cases, no further febrile episodes nor bacteraemia were observed and in two cases a complete positron emission tomography (PET) response was achieved. Due to a priori uncertainty about antimicrobial exposure during oral SAT, serum bactericidal titres (SBTs) were obtained and compared to those obtained during parenteral therapy. CONCLUSIONS: long term oral amoxicillin was effective and well-tolerated. SBTs after switch to oral therapy were quite heterogeneous, in some cases not reaching the conventionally established titre to assess bactericidal effect (1:8).Key pointsendovascular infection in non-suitable-for-surgery patients can be managed with long-term oral suppressive therapy.serum bactericidal assay confirmed high effectiveness of parenteral antibiotic therapy.serum bactericidal assay showed highly variable titres during oral therapy.

Is Advanced Age a Factor That Influences the Clinical Outcome of Single- or Double-Level MIS-TLIF? A Single-Center Study with a Minimum Two-Year Follow-Up on 103 Consecutive Cases.

As life expectancy rises, more elderly people undergo spinal fusion surgery to treat lumbar degenerative diseases. The MIS-TLIF technique, which minimizes soft tissue manipulation, is a promising fusion technique for frailer patients. The aim of this study was to investigate if older age is a significant factor in the clinical outcome of single- or double-level MIS-TLIF. A cross-sectional study was conducted on 103 consecutive patients. Data were compared between younger (<65 y.o.) and older (≥65 y.o.) patients. We observed no significant differences between baseline characteristics of the two groups apart from the frequency of disk space treated, with a relative predominance of L3-L4 space treated in the elderly (10% vs. 28%, p = 0.01) and L5-S1 space in younger patients (36% vs. 5%, p = 0.006). There was no significant difference in complication rate, surgical satisfaction, EQ 5D-5L, or Oswestry Disability Index (ODI) global or specific scores, with the exception of the EQ 5D-5L "mobility" score, where older patients fared worse (1.8 ± 1.1 vs. 2.3 ± 1.4; p = 0.05). The minimal invasiveness of the surgical technique, age-related specific outcome expectations, and biomechanical issues are all potential factors influencing the lack of age group differences in outcome scores.

Performance of an AI algorithm during the different phases of the COVID pandemics: what can we learn from the AI and vice versa.

Background: Artificial intelligence (AI) has proved to be of great value in diagnosing and managing Sars-Cov-2 infection. ALFABETO (ALL-FAster-BEtter-TOgether) is a tool created to support healthcare professionals in the triage, mainly in optimizing hospital admissions. Methods: The AI was trained during the pandemic's "first wave" (February-April 2020). Our aim was to assess the performance during the "third wave" of the pandemics (February-April 2021) and evaluate its evolution. The neural network proposed behavior (hospitalization vs home care) was compared with what was actually done. If there were discrepancies between ALFABETO's predictions and clinicians' decisions, the disease's progression was monitored. Clinical course was defined as "favorable/mild" if patients could be managed at home or in spoke centers and "unfavorable/severe" if patients need to be managed in a hub center. Results: ALFABETO showed accuracy of 76%, AUROC of 83%; specificity was 78% and recall 74%. ALFABETO also showed high precision (88%). 81 hospitalized patients were incorrectly predicted to be in "home care" class. Among those "home-cared" by the AI and "hospitalized" by the clinicians, 3 out of 4 misclassified patients (76.5%) showed a favorable/mild clinical course. ALFABETO's performance matched the reports in literature. Conclusions: The discrepancies mostly occurred when the AI predicted patients could stay at home but clinicians hospitalized them; these cases could be handled in spoke centers rather than hubs, and the discrepancies may aid clinicians in patient selection. The interaction between AI and human experience has the potential to improve both AI performance and our comprehension of pandemic management.

Editorial of Special Issues "Gut Microbiota-Host Interactions: From Symbiosis to Dysbiosis 2.0".

The gastrointestinal (GI) tract is where external agents meet the internal environment [...].

Dynamic NLR and PLR in Predicting COVID-19 Severity: A Retrospective Cohort Study.

INTRODUCTION: The hyperinflammation phase of severe SARS-CoV-2 is characterised by complete blood count alterations. In this context, the neutrophil-to-lymphocyte ratio (NLR) and the platelet-to-lymphocyte ratio (PLR) can be used as prognostic factors. We studied NLR and PLR trends at different timepoints and computed optimal cutoffs to predict four outcomes: use of continuous positive airways pressure (CPAP), intensive care unit (ICU) admission, invasive ventilation and death. METHODS: We retrospectively included all adult patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pneumonia admitted from 23 January 2020 to 18 May 2021. Analyses included non-parametric tests to study the ability of NLR and PLR to distinguish the patients' outcomes at each timepoint. Receiver operating characteristic (ROC) curves were built for NLR and PLR at each timepoint (minus discharge) to identify cutoffs to distinguish severe and non-severe disease. Their statistical significance was assessed with the chi-square test. Collection of data under the SMACORE database was approved with protocol number 20200046877. RESULTS: We included 2169 patients. NLR and PLR were higher in severe coronavirus disease 2019 (COVID-19). Both ratios were able to distinguish the outcomes at each timepoint. For NLR, the areas under the receiver operating characteristic curve (AUROC) ranged between 0.59 and 0.81, and for PLR between 0.53 and 0.67. From each ROC curve we computed an optimal cutoff value. CONCLUSION: NLR and PLR cutoffs are able to distinguish severity grades and mortality at different timepoints during the course of disease, and, as such, they allow a tailored approach. Future prospects include validating our cutoffs in a prospective cohort and comparing their performance against other COVID-19 scores.

Postvaccination immune-mediated hepatitis: what do we really know?

Tweetable abstract The percentage of patients with immune-mediated vaccine-associated hepatitis is minimal compared with the number of patients vaccinated worldwide.

A worldwide survey on perioperative antimicrobial prophylaxis for heart transplantation: From theory to clinical practice.

BACKGROUND: The aim of our survey was to analyze the current attitudes toward antimicrobial prophylaxis in heart transplanting centers worldwide. METHODS: The survey was composed of a total of 50 questions, it consisted of four different sections as follows. The first section collected physicians' personal data and centers' general characteristics, second assessed the approach to patients colonized with multidrug-resistant organisms (MDROs), while the third consisted of the infection risk related to cardiovascular devices, and antimicrobial treatment data, the last focused on donor's colonization. RESULTS: A total of 56 answers from 26 different countries were collected, mostly from Europe (n = 30) and the USA (n = 16). A first-generation cephalosporin (58.9%) or a combination therapy with vancomycin (10.7%) were the most frequently prescribed antimicrobial prophylaxis. Roughly 30% of the centers used different antimicrobial prophylaxis,mostly including Gram negative bacteria coverage. The frequency of screening for multidrug resistant Gram-negative bacteria was higher in Europe, where the percentage of centers providing screening for extended spectrum beta-lactamase (46.7%) and carbapenem-resistant Enterobacteriaceae (CRE) (53.3%) was higher than in other geographic area (p = .019; p = .013, respectively). CONCLUSION: This survey highlights a heterogeneity of clinical practice concerning antimicrobial prophylaxis at transplant. The concern for potential Gram-negative bacteria infection was responsible for broader antimicrobial coverage in 30% of centers.

Implementation of an antimicrobial stewardship program in the Vascular Surgery ward of a university tertiary care hospital in Pavia, Northern Italy.

PURPOSE: The commitment of multidisciplinary teams in antimicrobial stewardship programs (ASPs) is often inadequately considered, especially in surgical wards. We wanted to evaluate clinical, microbiological, and pharmacological outcomes before and after the implementation of an ASP in the Vascular Surgery ward of Fondazione IRCCS Policlinico San Matteo, a tertiary care hospital in Pavia, Italy. METHODS: This was a quasi-experimental quality-improvement study. The antimicrobial stewardship activity was conducted twice a week for 12 months and consisted of both prospective audit and feedback of all the ongoing antimicrobial prescriptions by the infectious diseases' consultants and educational meetings for the healthcare workers of the Vascular Surgery ward. For comparison between the study periods, Student t test (Mann-Whitney test for skewed distributions) was used for quantitative variables (ANOVA or Kruskall-Wallis for > 2 groups respectively), and Pearson's chi-squared test (Fisher exact test where appropriate) for categorical variables. 2-tailed tests were used. P-value significance cut-off was 0.05. RESULTS: During the 12-month intervention period, among a total number of 698 patients, 186 prescriptions were revised, mostly leading to de-escalating an ongoing antimicrobial therapy (39, 20.97%). A statistically significant reduction in isolates of carbapenem-resistant Pseudomonas aeruginosa (p-value 0.003) and the absence of Clostridioides difficile infections were reported. No statistically significant changes were observed in terms of length of stay and all-cause in-hospital mortality. A significant decrease in the administration of carbapenems (p-value 0.01), daptomycin (p-value < 0.01) and linezolid (p-value 0.43) was registered. A significant reduction in antimicrobial costs was also observed. CONCLUSIONS: The implementation of a 12-month ASP brought significant clinical and economic results, highlighting the benefits of a multidisciplinary teamwork.

Reduction of BSI associated mortality after a sepsis project implementation in the ER of a tertiary referral hospital.

The emergency room (ER) is the first gateway for patients with sepsis to inpatient units, and identifying best practices and benchmarks to be applied in this setting might crucially result in better patient's outcomes. In this study, we want to evaluate the results in terms of decreased the in-hospital mortality of patients with sepsis of a Sepsis Project developed in the ER. All patients admitted to the ER of our Hospital from the 1st January, 2016 to the 31stJuly 2019 with suspect of sepsis (MEWS score ≥ of 3) and positive blood culture upon ER admission were included in this retrospective observational study. The study comprises of two periods: Period A: From the 1st Jan 2016 to the 31st Dec 2017, before the implementation of the Sepsis project. Period B: From the 1st Jan 2018 to the 31stJul 2019, after the implementation of the Sepsis project. To analyze the difference in mortality between the two periods, a univariate and multivariate logistic regression was used. The risk of in-hospital mortality was expressed as an odds ratio (OR) and a 95% confidence interval (95% CI). Overall, 722 patients admitted in ER had positive BC on admissions, 408 in period A and 314 in period B. In-hospital mortality was 18.9% in period A and 12.7% in period B (p = 0.03). At multivariable analysis, mortality was still reduced in period B compared to period A (OR 0.64, 95% CI 0.41-0.98; p = 0.045). Having an infection due to GP bacteria or polymicrobial was associated with an increased risk of death, as it was having a neoplasm or diabetes. A marked reduction in in-hospital mortality of patients with documented BSI associated with signs or symptoms of sepsis after the implementation of a sepsis project based on the application of sepsis bundles in the ER.