Abstract Heart, Lung, Blood, and Sleep (HLBS) conditions such as heart failure, COPD, and pneumonia are among the most common causes of hospitalizations. Many of such hospitalizations are thought to be avoidable with early recognition and intervention. Today, patient-reported symptoms represent the primary means of detecting impending hospitalizations, but because symptoms are late indicators of disease, this results in days to weeks of delay in receiving care. We developed a non-contact adherence-independent longitudinal bed-sensing platform to detect early physiologic signs of impending hospitalizations before patients recognize or self-report symptoms. Preliminary results from our bed-based mechanical sensors suggest that the technology can learn patient-specific baselines during clinical stability and can recognize excursions from baseline in the early stages of developing illness. Here, we propose a milestone-based project aimed at collecting human training datasets and developing generalizable models that can recognize impending hospitalizations in the home with a focus on the underrepresented populations of San Diego and Imperial Valley counties. We now propose a administrative supplement to explore its ability to detect ventricular ectopy and arrhythmias. Doing so will encourage utilization and adoption across HLBS disease verticals and patient subpopulations.