A machine learning based exploration of COVID-19 mortality risk.

Early prediction of patient mortality risks during a pandemic can decrease mortality by assuring efficient resource allocation and treatment planning. This study aimed to develop and compare prognosis prediction machine learning models based on invasive laboratory and noninvasive clinical and demographic data from patients' day of admission. Three Support Vector Machine (SVM) models were developed and compared using invasive, non-invasive, and both groups. The results suggested that non-invasive features could provide mortality predictions that are similar to the invasive and roughly on par with the joint model. Feature inspection results from SVM-RFE and sparsity analysis displayed that, compared with the invasive model, the non-invasive model can provide better performances with a fewer number of features, pointing to the presence of high predictive information contents in several non-invasive features, including SPO2, age, and cardiovascular disorders. Furthermore, while the invasive model was able to provide better mortality predictions for the imminent future, non-invasive features displayed better performance for more distant expiration intervals. Early mortality prediction using non-invasive models can give us insights as to where and with whom to intervene. Combined with novel technologies, such as wireless wearable devices, these models can create powerful frameworks for various medical assignments and patient triage.

LV Dyssynchrony Assessed With Phase Analysis on Gated Myocardial Perfusion SPECT Can Predict Response to CRT in Patients With End-Stage Heart Failure.

BACKGROUND: Cardiac resynchronization therapy (CRT) is an established treatment in patients with end-stage heart failure and wide QRS complex. However, about 30% of patients do not benefit from CRT (non-responder). Recent studies with tissue Doppler imaging yielded disappointing results in predicting CRT responders. Phase analysis was developed to allow assessment of LV dyssynchrony by gated single photon emission computed tomography (SPECT) myocardial perfusion imaging (GMPS). OBJECTIVES: The aim of present study was to investigate the role of quantitative GMPS-derived LV dyssynchrony data to predict CRT responder. PATIENTS AND METHODS: Thirty eligible patients for CRT implantation underwent GMPS and echocardiography. Response to CRT was evaluated six months after the device implantation. Clinical response to CRT was defined as 50 meters increase in 6-minute walking test (6-MWT) distance. Echocardiographic response to CRT was defined as ≥ 15% decrease in left ventricular end-systolic volume (LVESV). The lead position was considered concordant if it was positioned at the area of latest mechanical activation, and discordant if located outside the area of latest mechanical activation. RESULTS: Clinical response to CRT was observed in 74% of patients. However, only 57% of patients were responder according to the echo criteria. There were statistically significant differences between CRT responders and non-responders for GMPS-derived variables, including phased histogram bandwidth (PHB), phase SD (PSD), and Entropy. Moreover, a cutoff value of 112° for PHB with a sensitivity of 72% and specificity of 70%, a cutoff value of 21° for PSD with a sensitivity of 90% and specificity of 74%, and a cutoff of 52% for Entropy with a sensitivity of 90% and a specificity of 80% were considered to discriminate responders and non-responders. CRT response was more likely in patients with concordant LV lead position compared to those with discordant LV lead position. CONCLUSIONS: GMPS-derived LV dyssynchrony variables can predict response to CRT with good sensitivity and specificity.