Multivariate Models of Blood Glucose Prediction in Type1 Diabetes: A Survey of the State-of-the-art.

Diabetes mellitus is a long-term chronicle disorder with a high prevalence rate worldwide. Continuous blood glucose and lifestyle monitoring enabled the control of blood glucose dynamics through machine learning applications using data created by various popular sensors. This survey aims to assess various classical time series, neural networks and state-of-the-art regression models based on a wide variety of machine learning techniques to predict blood glucose and hyper/hypoglycemia in Type 1 diabetic patients. The analysis covers blood glucose prediction modeling, regression, hyper/hypoglycemia alerts, diabetes diagnosis, monitoring, and management. However, the primary focus is on evaluating models for the prediction of Type 1 diabetes. A wide variety of machine learning algorithms have been explored to implement precision medicine by clinicians and provide patients with an early warning system. The automated pancreas may benefit from predictions and alerts of hyper and hypoglycemia.

Reengineering the Discharge Transition Process of COVID-19 Patients Using Telemedicine, Remote Patient Monitoring, and Around-the-Clock Remote Patient Monitoring from the Emergency Department and Inpatient Units.

Background: At the beginning of the COVID-19 pandemic, New York City quickly became the epicenter with hospitals at full capacity needing to care for patients. At New York Presbyterian Brooklyn Methodist Hospital, we needed to develop an innovative system of how to safely discharge the massive influx of patients. Inundation of patient care with limited manpower and resources forced us to align with a third-party vendor, around-the-clock alert, to make remote patient monitoring (RPM) possible. Each patient was prescribed a pulse oximeter and nurses were assigned to monitor vital signs, speak to patients, and escalate to physicians if required. Results: We enrolled 50 patients, of whom 13 were escalated resulting in 3 emergency room visits and 1 readmission. We had a high compliance rate with high patient satisfaction in postsurveys. Discussion: Our program was unique in that it utilized telemedicine for regular patient follow-up, along with RPM through a third-party vendor. Patients were able to be safely discharged home with close follow-up through regularly obtained vitals with access to a 24/7 hotline for any emergencies, possibly preventing readmissions. Limitations include a small sample size population. Conclusions: Our experience shows that in a short period despite lack of resources, telehealth and RPM's concurrent use with a third-party vendor could be successfully utilized for safe discharges with high patient satisfaction.