Cost-sensitive ordinal classification methods to predict SARS-CoV-2 pneumonia severity.

OBJECTIVE: To study the suitability of costsensitive ordinal artificial intelligence-machine learning (AIML) strategies in the prognosis of SARS-CoV-2 pneumonia severity. MATERIALS & METHODS: Observational, retrospective, longitudinal, cohort study in 4 hospitals in Spain. Information regarding demographic and clinical status was supplemented by socioeconomic data and air pollution exposures. We proposed AI-ML algorithms for ordinal classification via ordinal decomposition and for cost-sensitive learning via resampling techniques. For performance-based model selection, we defined a custom score including per-class sensitivities and asymmetric misprognosis costs. 260 distinct AI-ML models were evaluated via 10 repetitions of 5×5 nested cross-validation with hyperparameter tuning. Model selection was followed by the calibration of predicted probabilities. Final overall performance was compared against five well-established clinical severity scores and against a 'standard' (non-cost sensitive, non-ordinal) AI-ML baseline. In our best model, we also evaluated its explainability with respect to each of the input variables. RESULTS: The study enrolled n = 1548 patients: 712 experienced low, 238 medium, and 598 high clinical severity. d = 131 variables were collected, becoming d ' = 148 features after categorical encoding. Model selection resulted in our best-performing AI-ML pipeline having: a) no imputation of missing data, b) no feature selection (i.e. using the full set of d ' features), c) 'Ordered Partitions' ordinal decomposition, d) cost-based reimbalance, and e) a Histogram-based Gradient Boosting classifier. This best model (calibrated) obtained a median accuracy of 68.1% [67.3%, 68.8%] (95% confidence interval), a balanced accuracy of 57.0% [55.6%, 57.9%], and an overall area under the curve (AUC) 0.802 [0.795, 0.808]. In our dataset, it outperformed all five clinical severity scores and the 'standard' AI-ML baseline. DISCUSSION & CONCLUSION: We conducted an exhaustive exploration of AI-ML methods designed for both ordinal and cost-sensitive classification, motivated by a real-world application domain (clinical severity prognosis) in which these topics arise naturally. Our model with the best classification performance exploited successfully the ordering information of ground truth classes, coping with imbalance and asymmetric costs. However, these ordinal and cost-sensitive aspects are seldom explored in the literature.

Extracting relevant predictive variables for COVID-19 severity prognosis: An exhaustive comparison of feature selection techniques.

With the COVID-19 pandemic having caused unprecedented numbers of infections and deaths, large research efforts have been undertaken to increase our understanding of the disease and the factors which determine diverse clinical evolutions. Here we focused on a fully data-driven exploration regarding which factors (clinical or otherwise) were most informative for SARS-CoV-2 pneumonia severity prediction via machine learning (ML). In particular, feature selection techniques (FS), designed to reduce the dimensionality of data, allowed us to characterize which of our variables were the most useful for ML prognosis. We conducted a multi-centre clinical study, enrolling n = 1548 patients hospitalized due to SARS-CoV-2 pneumonia: where 792, 238, and 598 patients experienced low, medium and high-severity evolutions, respectively. Up to 106 patient-specific clinical variables were collected at admission, although 14 of them had to be discarded for containing ⩾60% missing values. Alongside 7 socioeconomic attributes and 32 exposures to air pollution (chronic and acute), these became d = 148 features after variable encoding. We addressed this ordinal classification problem both as a ML classification and regression task. Two imputation techniques for missing data were explored, along with a total of 166 unique FS algorithm configurations: 46 filters, 100 wrappers and 20 embeddeds. Of these, 21 setups achieved satisfactory bootstrap stability (⩾0.70) with reasonable computation times: 16 filters, 2 wrappers, and 3 embeddeds. The subsets of features selected by each technique showed modest Jaccard similarities across them. However, they consistently pointed out the importance of certain explanatory variables. Namely: patient's C-reactive protein (CRP), pneumonia severity index (PSI), respiratory rate (RR) and oxygen levels -saturation Sp O2, quotients Sp O2/RR and arterial Sat O2/Fi O2-, the neutrophil-to-lymphocyte ratio (NLR) -to certain extent, also neutrophil and lymphocyte counts separately-, lactate dehydrogenase (LDH), and procalcitonin (PCT) levels in blood. A remarkable agreement has been found a posteriori between our strategy and independent clinical research works investigating risk factors for COVID-19 severity. Hence, these findings stress the suitability of this type of fully data-driven approaches for knowledge extraction, as a complementary to clinical perspectives.

An NMR-Based Model to Investigate the Metabolic Phenoreversion of COVID-19 Patients throughout a Longitudinal Study.

After SARS-CoV-2 infection, the molecular phenoreversion of the immunological response and its associated metabolic dysregulation are required for a full recovery of the patient. This process is patient-dependent due to the manifold possibilities induced by virus severity, its phylogenic evolution and the vaccination status of the population. We have here investigated the natural history of COVID-19 disease at the molecular level, characterizing the metabolic and immunological phenoreversion over time in large cohorts of hospitalized severe patients (n = 886) and non-hospitalized recovered patients that self-reported having passed the disease (n = 513). Non-hospitalized recovered patients do not show any metabolic fingerprint associated with the disease or immune alterations. Acute patients are characterized by the metabolic and lipidomic dysregulation that accompanies the exacerbated immunological response, resulting in a slow recovery time with a maximum probability of around 62 days. As a manifestation of the heterogeneity in the metabolic phenoreversion, age and severity become factors that modulate their normalization time which, in turn, correlates with changes in the atherogenesis-associated chemokine MCP-1. Our results are consistent with a model where the slow metabolic normalization in acute patients results in enhanced atherosclerotic risk, in line with the recent observation of an elevated number of cardiovascular episodes found in post-COVID-19 cohorts.

Diagnostic and therapeutic approach to occupational pneumonia

We shall define occupational pneumonia as a diseaseof external origin, closely tied to the workplace setting andcaused by biological microorganisms. The main pathogens arebacteria, fungi and viruses. There are a number of occupationsspecifically prone to the possibility of acquiring pneumoniawhen performing work duties.In addition to the diagnostic methods and drug treatments current in infectious processes, a good clinical history,with avoidance and protection measures would be the mostimportant tools for the management of occupational pneumonia.Social and demographic changes in the last two decadeshave made zoonotic infections, and especially viruses, the maincause of new infections. Human health and animal health areclosely linked, so collaboration between veterinarians and doctors, together with the necessary environmental respect andconservation, plus the appropriate public policies are essentialto avoid these wide negative effects. (AU)

Bacteraemic pneumococcal pneumonia and SARS-CoV-2 pneumonia: differences and similarities.

OBJECTIVE: To analyse differences in clinical presentation and outcome between bacteraemic pneumococcal community-acquired pneumonia (B-PCAP) and sSvere Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) pneumonia. METHODS: This observational multi-centre study was conducted on patients hospitalized with B-PCAP between 2000 and 2020 and SARS-CoV-2 pneumonia in 2020. Thirty-day survival, predictors of mortality, and intensive care unit (ICU) admission were compared. RESULTS: In total, 663 patients with B-PCAP and 1561 patients with SARS-CoV-2 pneumonia were included in this study. Patients with B-PCAP had more severe disease, a higher ICU admission rate and more complications. Patients with SARS-CoV-2 pneumonia had higher in-hospital mortality (10.8% vs 6.8%; P=0.004). Among patients admitted to the ICU, the need for invasive mechanical ventilation (69.7% vs 36.2%; P<0.001) and mortality were higher in patients with SARS-CoV-2 pneumonia. In patients with B-PCAP, the predictive model found associations between mortality and systemic complications (hyponatraemia, septic shock and neurological complications), lower respiratory reserve and tachypnoea; chest pain and purulent sputum were protective factors in these patients. In patients with SARS-CoV-2 pneumonia, mortality was associated with previous liver and cardiac disease, advanced age, altered mental status, tachypnoea, hypoxaemia, bilateral involvement, pleural effusion, septic shock, neutrophilia and high blood urea nitrogen; in contrast, ≥7 days of symptoms was a protective factor in these patients. In-hospital mortality occurred earlier in patients with B-PCAP. CONCLUSIONS: Although B-PCAP was associated with more severe disease and a higher ICU admission rate, the mortality rate was higher for SARS-CoV-2 pneumonia and deaths occurred later. New prognostic scales and more effective treatments are needed for patients with SARS-CoV-2 pneumonia.

Proteína C reactiva, procalcitonina y proadrenomedulina en la evolución de neumonías hospitalizadas / C reactive protein, procalcitonin and proadrenomedullin in the outcome of hospitalized pneumonia patients

Introducción. La neumonía adquirida en la comunidad (NAC) sigue siendo un problema sanitario importante. Para establecer su gravedad existen una serie de escalas de severidad, pero tienen sus limitaciones. Se han propuesto diferentes biomarcadores que podrían resultar de ayuda. Objetivo. Evaluar el valor pronóstico de proteína C reactiva (PCR), procalcitonina (PCT) y proadrenomedulina (PADM) para predecir mala evolución intrahospitalaria en NAC. Material y métodos. Se incluyeron todos los pacientes diagnosticados de NAC que quedaron ingresados durante un periodo de 13 meses. Se congeló a −80°C suero y plasma EDTA obtenidos en el Servicio de Urgencias del Hospital para la determinación de los biomarcadores. Se dividió a los pacientes en dos grupos: los que evolucionaron favorablemente y los que tuvieron mala evolución. Los datos clínicos de los pacientes fueron recopilados por revisión de la historia clínica. Resultados. Las diferencias de las medianas de los tres biomarcadores para los dos grupos adquirieron significación estadística. Las áreas bajo la curva de las curvas ROC correspondientes fueron: 0,67 para PCT, 0,62 para PCR y 0,74 para PADM. Los puntos de corte seleccionados con sus respectivos datos de sensibilidad y especificidad fueron: para PCT 0,5 ng/mL (S: 0,67/E: 0,61), para PCR 150mg/L (S: 0,67/E: 0,47) y para PADM 1,2 nmol/L (S: 0,80/E: 0,53). Conclusiones. Los resultados sugieren un posible valor pronóstico de estos biomarcadores en relación con la evolución intrahospitalaria que presentarán los pacientes con NAC, destacando entre ellos la PADM (AU)

Introduction: Community-acquired pneumonia (CAP) continues to be a major health problem. There are several scoring systems to predict its severity, but they have limitations. Different biomarkers have been proposed to be of assistance. Objective: To evaluate C reactive protein (CRP), procalcitonin (PCT) and proadrenomedullin (PADM) as prognostic factors to predict the outcome in CAP. Material and methods: All patients diagnosed with CAP and admitted to hospital during a period of 13 months were included in our study. Serum and EDTA plasma samples from the Emergency Unit were collected and frozen at -80 ◦C for biomarkers determination. Patients were divided into two groups: those who developed favorably and those with an unfavorable outcome. Clinical data for these patients were collected by reviewing their medical records. Results: The median values between both groups were found to be statistically significantly different for all three biomarkers. Areas under the ROC curve for each biomarker were: 0.67 for PCT, 0.62 for CRP and 0.74 for PADM. Selected cut-off for each biomarker with their corresponding sensitivity and specificity values were: 0.5 ng/mL (Se: 0.67/Sp: 0.61) for PCT, 150 mg/L (Se: 0.67/Sp: 0.47) for CRP and 1.2 nmol/L (Se: 0.8/Sp: 0.53) for PADM. Conclusions: The results indicate that these biomarkers could help in predicting the outcome of patients with CAP during hospitalization, with PADM being a potentially better predictor (AU)