Search

Background: Several models have been developed to predict mortality in patients with COVID-19 pneumonia, but only a few have demonstrated enough discriminatory capacity. Machine learning algorithms represent a novel approach for the data-driven prediction of clinical outcomes with advantages over statistical modeling., Objective: We aimed to develop a machine learning-based score-the Piacenza score-for 30-day mortality prediction in patients with COVID-19 pneumonia., Methods: The study comprised 852 patients with COVID-19 pneumonia, admitted to the Guglielmo da Saliceto Hospital in Italy from February to November 2020. Patients' medical history, demographics, and clinical data were collected using an electronic health record. The overall patient data set was randomly split into derivation and test cohorts. The score was obtained through the naïve Bayes classifier and externally validated on 86 patients admitted to Centro Cardiologico Monzino (Italy) in February 2020. Using a forward-search algorithm, 6 features were identified: age, mean corpuscular hemoglobin concentration, PaO 2 /FiO 2 ratio, temperature, previous stroke, and gender. The Brier index was used to evaluate the ability of the machine learning model to stratify and predict the observed outcomes. A user-friendly website was designed and developed to enable fast and easy use of the tool by physicians. Regarding the customization properties of the Piacenza score, we added a tailored version of the algorithm to the website, which enables an optimized computation of the mortality risk score for a patient when some of the variables used by the Piacenza score are not available. In this case, the naïve Bayes classifier is retrained over the same derivation cohort but using a different set of patient characteristics. We also compared the Piacenza score with the 4C score and with a naïve Bayes algorithm with 14 features chosen a priori., Results: The Piacenza score exhibited an area under the receiver operating characteristic curve (AUC) of 0.78 (95% CI 0.74-0.84, Brier score=0.19) in the internal validation cohort and 0.79 (95% CI 0.68-0.89, Brier score=0.16) in the external validation cohort, showing a comparable accuracy with respect to the 4C score and to the naïve Bayes model with a priori chosen features; this achieved an AUC of 0.78 (95% CI 0.73-0.83, Brier score=0.26) and 0.80 (95% CI 0.75-0.86, Brier score=0.17), respectively., Conclusions: Our findings demonstrated that a customizable machine learning-based score with a purely data-driven selection of features is feasible and effective for the prediction of mortality among patients with COVID-19 pneumonia., (©Geza Halasz, Michela Sperti, Matteo Villani, Umberto Michelucci, Piergiuseppe Agostoni, Andrea Biagi, Luca Rossi, Andrea Botti, Chiara Mari, Marco Maccarini, Filippo Pura, Loris Roveda, Alessia Nardecchia, Emanuele Mottola, Massimo Nolli, Elisabetta Salvioni, Massimo Mapelli, Marco Agostino Deriu, Dario Piga, Massimo Piepoli. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 31.05.2021.)

To date, the pandemic spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has involved over 100 countries in a matter of weeks, and Italy suffers from almost 1/3 of the dead cases worldwide. In this report, we show the strategies adopted to face the emergency at Centro Cardiologico Monzino, a mono-specialist cardiology hospital sited in the region of Italy most affected by the pandemic, and specifically we describe how we have progressively modified in a few weeks the organization of our Heart Failure Unit in order to cope with the new COVID-19 outbreak. In fact, on the background of the pandemic, cardiovascular diseases still occur frequently in the general population, but we observed consistent reduction in hospital admissions for acute cardiovascular events and a dramatic increase of late presentation acute myocardial infarction. Despite a reduction of healthcare workers number, our ward has been rearranged in order to take care of both COVID-19 and cardiovascular patients. In particular according to a triple step procedure we divided admitted patients in confirmed, suspected and excluded cases (respectively allocated in "red", "pink" and "green" separated areas). Due to the absence of definite guidelines, our aim was to describe our strategy in facing the current emergency, in order to reorganize our hospital in a dynamic and proactive manner. To quote the famous Italian writer Alessandro Manzoni 'It is less bad to be agitated in doubt than to rest in error.'

Aims: Ventilation vs. carbon dioxide production (VE/VCO 2 ) is among the strongest cardiopulmonary exercise testing prognostic parameters in heart failure (HF). It is usually reported as an absolute value. The current definition of normal VE/VCO 2 slope values is inadequate, since it was built from small groups of subjects with a particularly limited number of women and elderly. We aimed to define VE/VCO 2 slope prediction formulas in a sizable population and to test whether the prognostic power of VE/VCO 2 slope in HF was different if expressed as a percentage of the predicted value or as an absolute value., Methods and Results: We calculated the linear regressions between age and VE/VCO 2 slope in 1136 healthy subjects (68% male, age 44.9 ± 14.5, range 13-83 years). We then applied age-adjusted and sex-adjusted formulas to predict VE/VCO 2 slope to HF patients included in the metabolic exercise test data combined with cardiac and kidney indexes score database, which counts 6112 patients (82% male, age 61.4 ± 12.8, left ventricular ejection fraction 33.2 ± 10.5%, peakVO 2 14.8 ± 4.9, mL/min/kg, VE/VCO 2 slope 32.7 ± 7.7) from 24 HF centres. Finally, we evaluated whether the use of absolute values vs. percentages of predicted VE/VCO 2 affected HF prognosis prediction (composite of cardiovascular mortality + urgent transplant or left ventricular assist device). We did so in the entire cardiac and kidney indexes score population and separately in HF patients with severe (peakVO 2 < 14 mL/min/kg, n = 2919, 61.1 events/1000 pts/year) or moderate (peakVO 2 ≥ 14 mL/min/kg, n = 3183, 19.9 events/1000 pts/year) HF. In the healthy population, we obtained the following equations: female, VE/VCO 2 = 0.052 × Age + 23.808 (r = 0.192); male, VE/VCO 2 = 0.095 × Age + 20.227 (r = 0.371) (P = 0.007). We applied these formulas to calculate the percentages of predicted VE/VCO 2 values. The 2-year survival prognostic power of VE/VCO 2 slope was strong, and it was similar if expressed as absolute value or as a percentage of predicted value (AUCs 0.686 and 0.690, respectively). In contrast, in severe HF patients, AUCs significantly differed between absolute values (0.637) and percentages of predicted values (0.650, P = 0.0026). Moreover, VE/VCO 2 slope expressed as a percentage of predicted value allowed to reclassify 6.6% of peakVO 2 < 14 mL/min/kg patients (net reclassification improvement = 0.066, P = 0.0015)., Conclusions: The percentage of predicted VE/VCO 2 slope value strengthens the prognostic power of VE/VCO 2 in severe HF patients, and it should be preferred over the absolute value for HF prognostication. Furthermore, the widespread use of VE/VCO 2 slope expressed as percentage of predicted value can improve our ability to identify HF patients at high risk, which is a goal of utmost clinical relevance., (© 2020 The Authors. ESC Heart Failure published by John Wiley & Sons Ltd on behalf of the European Society of Cardiology.)

Aims: Exercise oscillatory ventilation (EOV) is a pivotal cardiopulmonary exercise test parameter for the prognostic evaluation of patients with chronic heart failure (HF). It has been described in patients with HF with reduced ejection fraction (<40%, HFrEF) and with HF with preserved ejection fraction (>50%, HFpEF), but no data are available for patients with HF with mid-range ejection fraction (40-49%, HFmrEF). The aim of the study was to evaluate the prognostic role of EOV in HFmrEF patients., Methods and Results: We analysed 1239 patients with HFmrEF and 4482 patients with HFrEF, enrolled in the MECKI score database, with a 2-year follow-up. The study endpoint was the composite of cardiovascular death, urgent heart transplant, and ventricular assist device implantation. We identified EOV in 968 cases (16% and 17% of cases in HFmrEF and HFrEF, respectively). HFrEF EOV+ patients were significantly older, and their parameters suggested a more severe HF than HFrEF EOV- patients. A similar behaviour was found in HFmrEF EOV+ vs. EOV- patients. Kaplan-Meier analysis, irrespective of ejection fraction, showed that EOV is associated with a worse survival, and that patients with HFrEF and HFmrEF EOV+ had a significantly worse outcome than the EOV- of the same ejection fraction groups. EOV-associated survival differences in HFmrEF patients started after 18 months of follow-up., Conclusion: Exercise oscillatory ventilation has a similar prevalence and ominous prognostic value in both HFmrEF and HFrEF patients, indicating a group of patients in need of a more intensive follow-up and a more aggressive therapy. In HFmrEF, the survival curves between EOV+ and EOV- patients diverged only after 18 months., (© 2019 The Authors. European Journal of Heart Failure © 2019 European Society of Cardiology.)

Aims: Risk stratification in heart failure (HF) is crucial for clinical and therapeutic management. A multiparametric approach is the best method to stratify prognosis. In 2012, the Metabolic Exercise test data combined with Cardiac and Kidney Indexes (MECKI) score was proposed to assess the risk of cardiovascular mortality and urgent heart transplantation. The aim of the present study was to compare the prognostic accuracy of MECKI score to that of HF Survival Score (HFSS) and Seattle HF Model (SHFM) in a large, multicentre cohort of HF patients with reduced ejection fraction., Methods and Results: We collected data on 6112 HF patients and compared the prognostic accuracy of MECKI score, HFSS, and SHFM at 2- and 4-year follow-up for the combined endpoint of cardiovascular death, urgent cardiac transplantation, or ventricular assist device implantation. Patients were followed up for a median of 3.67 years, and 931 cardiovascular deaths, 160 urgent heart transplantations, and 12 ventricular assist device implantations were recorded. At 2-year follow-up, the prognostic accuracy of MECKI score was significantly superior [area under the curve (AUC) 0.781] to that of SHFM (AUC 0.739) and HFSS (AUC 0.723), and this relationship was also confirmed at 4 years (AUC 0.764, 0.725, and 0.720, respectively)., Conclusion: In this cohort, the prognostic accuracy of the MECKI score was superior to that of HFSS and SHFM at 2- and 4-year follow-up in HF patients in stable clinical condition. The MECKI score may be useful to improve resource allocation and patient outcome, but prospective evaluation is needed., (© 2017 The Authors. European Journal of Heart Failure © 2017 European Society of Cardiology.)

Background: In patients with heart failure (HF), during maximal cardiopulmonary exercise test, anaerobic threshold (AT) is not always identified. We evaluated whether this finding has a prognostic meaning., Methods and Results: We recruited and prospectively followed up, in 14 dedicated HF units, 3058 patients with systolic (left ventricular ejection fraction <40%) HF in stable clinical conditions, New York Heart Association class I to III, who underwent clinical, laboratory, echocardiographic, and cardiopulmonary exercise test investigations at study enrollment. We excluded 921 patients who did not perform a maximal exercise, based on lack of achievement of anaerobic metabolism (peak respiratory quotient ≤1.05). Primary study end point was a composite of cardiovascular death and urgent cardiac transplant, and secondary end point was all-cause death. Median follow-up was 3.01 (1.39-4.98) years. AT was identified in 1935 out of 2137 patients (90.54%). At multivariable logistic analysis, failure in detecting AT resulted significantly in reduced peak oxygen uptake and higher metabolic exercise and cardiac and kidney index score value, a powerful prognostic composite HF index (P<0.001). At multivariable analysis, the following variables were significantly associated with primary study end point: peak oxygen uptake (% pred; P<0.001; hazard ratio [HR]=0.977; confidence interval [CI]=0.97-0.98), ventilatory efficiency slope (P=0.01; HR=1.02; CI=1.01-1.03), hemoglobin (P<0.05; HR=0.931; CI=0.87-1.00), left ventricular ejection fraction (P<0.001; HR=0.948; CI=0.94-0.96), renal function (modification of diet in renal disease; P<0.001; HR=0.990; CI=0.98-0.99), sodium (P<0.05; HR=0.967; CI=0.94-0.99), and AT nonidentification (P<0.05; HR=1.41; CI=1.06-1.89). Nonidentification of AT remained associated to prognosis also when compared with metabolic exercise and cardiac and kidney index score (P<0.01; HR=1.459; CI=1.09-1.10). Similar results were obtained for the secondary study end point., Conclusions: The inability to identify AT most often occurs in patients with severe HF, and it has an independent prognostic role in HF.