IoT-Enabled Healthcare Management: Enhancing Patient Monitoring and Service Delivery Smart Healthcare Management Through Internet of Things Technologies

This is the authors' abstract. We don't add key points for this paper.

We Continuous physiological monitoring through healthcare system with Internet of Things (IoT) technologies is useful for early detection of any impending instability, however, the reliability, strong communication, efficient computation and integration into workflows of such systems are all key factors for their effectiveness. This study created an IoT-based machine learning framework for future next window physiological-instability assessment and healthcare services prioritization using the BIDMC PPG and Respiration Dataset that contains data from 53 subjects. Physiological sensors used were heart rate, pulse, respiratory rate, SpO₂, respiration waveform, photoplethysmography and ECG Lead II. Five fold subject-grouped cross validation was performed for six machine-learning models. Random Forest achieved the strongest overall performance, with accuracy of 0.8706, sensitivity of 0.8061, specificity of 0.9223, F1-score of 0.8471, ROC-AUC of 0.9366, PR-AUC of 0.9110, and MCC of 0.7384. Model stability was confirmed through bootstrap and calibration analyses, and permutation importance showed that oxygen-saturation, heart-rate and respiratory features were important. Preserved discrimination was demonstrated through robustness testing in the case of packet loss and sensor noise. Favorable latency-bandwidth trade-offs and 38.0% reduction in mean waiting time for future-unstable episodes was further demonstrated in simulated Edge–Cloud–Hybrid deployments, as well as risk-prioritized scheduling. Clinical validation (outside the lab) is still required prior to deployment in the real world.