Machine learning for prediction of in-hospital mortality in COVID-19 patients: Results from an Italian multicentre study

Aims: Several risk factors have been identified to predict worse outcomes in patients affected by SARS-CoV-2 infection. Prediction models are needed to optimize clinical management and to early stratify patients at a higher mortality risk. Machine learning (ML) algorithms represent a novel approach to identify a prediction model with a good discriminatory capacity to be easily used in clinical practice. Methods and results: The Cardio-COVID is a multicentre observational study that involved a cohort of consecutive adult Caucasian patients with laboratory-confirmed COVID-19 [by real time reverse transcriptase-polymerase chain reaction (RT-PCR)] who were hospitalized in 13 Italian cardiology units from 1 March to 9 April 2020. Patients were followed-up after the COVID-19 diagnosis and all causes in-hospital mortality or discharge were ascertained until 23 April 2020. Variables with more than 20% of missing values were excluded. The Lasso procedure was used with a λ=0.07 for reducing the covariates number. Mortality was estimated by means of a Random Forest (RF). The dataset was randomly divided in two subsamples with the same percentage of death/alive people of the entire sample: training set contained 80% of the data and test set the remaining 20%. The training set was used in the calibration procedure where a RF models in-hospital mortality with the covariates selected by Lasso. Its accuracy was measured by means of the ROC curve, obtaining AUC, sensitivity, specificity, and related 95% confidence interval (CI) computed with 10 000 stratified bootstrap replicates. From the RF the relative Variable Importance Measure (relVIM) was extracted to understand which of the selected variables had the greatest impact on outcome, providing a ranking from the most (relVIM=100) to the less important variable. The model obtained was compared with the Gradient Boosting Machine (GBM) and with the logistic regression, where the predictions were cross validated. Finally, to understand if each model has the same performance in sample (training) and out of sample (test), the two AUCs were compared by means of the DeLong's test. Among 701 patients enrolled (mean age 67.2±13.2 years, 69.5% males), 165 (23.5%) died during a median hospitalization of 15 (IQR, 9-24) days. Variables selected by the Lasso were: age, Oxygen saturation, PaO2/FiO2, Creatinine Clearance and elevated Troponin. Compared with those who survived, deceased patients were older, had a lower blood oxygenation, a lower creatinine clearance levels and higher prevalence of elevated Troponin (all P<0.001). Training set included 561 patients and test set 140 patients. The best performance out of sample was provided by the RF with an AUC of 0.78 (95% CI: 0.68-0.88) and a sensitivity of 0.88 (95% CI: 0.58-1.00). Moreover, RF is the unique methodology that provided similar performance in sample and out of sample (DeLong test P=0.78). On the contrary, prediction model was less accurate by using GBM and logistic regression. The relVIM ranked the variables from the most to the less important in predicting the outcome as follows: clearance creatinine, PaO2/FiO2, age, oxygen saturation, and elevated Troponin. Conclusions: In a large COVID-19 population, we showed that a customizable MLbased score derived from clinical variables, is feasible and effective for the prediction of in-hospital mortality.

Rehabilitation in post-COVID patients. A single center experience

The COVID 19 disease is frequently associated with significant disability related to intensive care unit-acquired weakness, decontitioning, myopathies and neuropathies. However there are no data on the results of a specific rehabilitative treatment in this group of patients. The aim of our work was to evaluate the effectiveness f a personalized rehabilitative therapy in group of post-COVID patients (A, 47 patients, average age 65.3±11.6 y, 27 M,) comparing the results with a group of postcardiosurgical patients COVID 19 negative (B, 47 patients, average age 63.5±10.3 y, 29 M) evaluating the degree of clinical complexity (Rehabilitation Complexity Scale, RCS-E V13) and the degree of autonomy recovery (Six-minute walking test SMWT, Barthel Index, BI) pre and post-treatment. In Group A patients the Rehabilitation program is associated with a significant improvement in autonomy recovery (BI admission 29.7±20 vs discharge 72.7±28.6 p<0.005, SMWT admission 146±25 vs 318±18 m, p<0.005) and in clinical complexity (RCS admission 10.9±1.1 vs discharge 5.3, p<0.05). At admission the comparison between Group A vs B has show: 1. a reduced pre-rehabilitation hospital stay (days) in Group Vs A (B 8.2±2 vs 31±5 0.005) 2. a similar degrre of clinical complexity (RCS scale A 10.9±1.1 vs 1.6±11.2 p ns) 3. a greater loss autonomy in post-COVID patients (BI scale A 29.7±20 vs B 47.7±19, p 0.05;SMWT A 145±25 m vs B 255±18 m, p 0.05) After a similar period of rehabilitation (A 29.7±12.8 days vs B 29.6±10 days, p ns) we observed in both Groups: 1. a reduction of clinical complexity ((RCS scale A 5.3±2 vs 6.6±2 p ns 2. an improvement of degree of autonomy recovery ((BI scale A 72.7±28 vs B 47.7±19, p ns;SMWT A 385±18 m vs B 410±25m, p ns) Conclusions: Post-COVID patients show a greater loss of autonomy than post-cardiosurgery patients. Rehabilitative treatment has proven effective in ensuring adequate functional recovery with similar results to those obtained in the population of cardiological subjects COVID 19 negative.

Utility of remote monitoring for implantable device controls in the phase 1 of the first wave of COVID 19 pandemic

Backgrond: During the first outbreak of COVID 19 pandemic, Piacenza was particularly affected since it counted over one hundred daily access in emergency department, of patients (pz) with SARS COV2 virus. This fact led to a reorganization of hospital activity consisting in the formation of 7 department devoting COVID 19 care and in temporarily postponing scheduled activities of various disciplines to prevent the spread of virus. Also Our Division of cardiology has kept only urgent clinical and interventional activity. With the use of remote monitoring (RM), we were able to check implantable cardiac devices (ICD) almost scheduled and we called ICD recipients in office only for urgent troubles. Methods: In our study we evaluated all ICD recipients that had a scheduled follow up in our electrostimulation clinic on the period from 23th February 2020 to 18 th May 2020 Results: In office scheduled controls during the period considered, involved 216 patients. 85% out of them was followed also with MR;after postponing in office visits, we requested control transmissions. In total we received 441 scheduled and with alert transmissions. Regarding alert transmission: 3 of them signaled ERI (elective replacement indicator), so the replacement of device has been planned;3 of them indicated noise in Ventricular Fibrillation zone related to lead malfunction, so we planned reimplantation of new ventricular lead;12 recorded ventricular arrhythmias (only one patient was called to visit in office for recurrent ventricular arrhythmias);1 of them signaled long lasting atrial fibrillation so we called him to begin anticoagulant therapy. We performed phone triage before confirming in office visit. None of the scheduled transmissions detected troubles. Only 1,8% of patients followed by remote monitoring came in hospital in that period. Conclusions: RM during phase 1 of the first wave of COVID 19 outbreak allowed us to reduce in office visits and to call in hospital only patients with real needs, decreasing the spread of the virus and maintaining identification of clinical and technical troubles.

The usefulness of dipyridamole stress echocardiography in high-risk patients before abdominal aneurism surgery

Background: Coronary artery disease (CAD) and aortic aneurysm (AA) share commons risk factors, such as hypertension, diabetes mellitus, hypercholesterolemia, and smoking. Cardiac assessment before aortic abdominal aneurysm (AAA) surgery is indicated for patients with symptomatic coronary artery disease (CAD). The usefulness of assessment of moderate/high-risk patients is still debated. Purpose: the purpose of our study is to evaluate the safety and effectiveness of dipyridamole stress echocardiography (DSE) for the detection of CAD in patients undergoing AAA surgery with high cardiovascular risk. Methods: From 2017th to 2019th 120 patients underwent surgery for aortic aneurysm (71 endovascular technique and 49 with open laparot-omy). Of these, 74 asymptomatic patients with high cardiovascular risk underwent a pre-surgical contrast-enhanced dipyridamole stress echo (0,84 mg/kg over 6 minutes - protocol with LVO with sulfur hexafluoride), to exclude the presence of inducible myocardial ischemia, Mean follow-up was 6-24 months. Results: Mean age was 77 years +/- 6.6, with male gender prevalent (83%). No complication during DSE occurred;mean SCORE risk was 9.8% +/- 2.3%, with 63% patients with very high risk. Only 1 patient showed inducible ischemia during stress echocardiography, with evidence of significant LAD stenosis;no myocardial infarction was reported at follow-up, while 1 ischemic stroke and 1 unplanned revasculari-zation occurred. 11% of patients died, of which 50% for Sars-Cov-2 disease and 12% due to post-surgery dissection while no cardiac deaths were found. Conclusions: dipyridamole stress echo is safe in patients with surgical-class abdominal aortic aneurism;in patients with high cardiovascular risk but no symptoms reversible ischemia is rare. DSE should not be routinely performed before high-risk surgery but only in patients with cardiac symptoms.