Link-HF: Multisensor Non-invasive Telemonitoring System for Prediction of Heart Failure Exacerbation

The analytics being investigated includes a Similarity-Based Modeling technique, that empirically estimates the expected physiological behavior of a subject based on prior learned dynamic data, for comparison to actual measured behavior from the subject, to reveal discrepancies hidden by normal variation. The measurements are typically an ensemble of vital signs that effectively characterizes the physiological "control system" of the subject. This technique is multivariate: multiple variables are leveraged, because single variables in isolation have little context - a high heart rate by itself could mean a person is exerting himself, or it could mean his physiology is in distress even though he is not exerting himself. With reference to several other variables, however, such as respiration rate, oximetry and motion/activity, a high heart rate might be recognized as a normal state when accompanied by the corroborating data showing a high respiration rate, a normal oximetry and a high level of motion - the person is exercising. A wearable adhesive multi-sensor device will be used to collect continuous vital sign and other data from study subjects, including heart rate, respiration rate, bodily motion/activity, skin temperature, pulse, electrocardiogram and peripheral capillary oxygen saturation. Subjects are provided with a smartphone or cellular tablet that will be paired with the multi-sensor device to receive data and upload it to the analytics server via cellular network or WiFi internet. Study staff will interact with the subject during visits scheduled for routine heart failure follow-up to capture pre-specified heart failure medical events. All standard of care clinic and hospitalization notes and procedure reports including echocardiograms, right heart catheterizations, pulmonary function tests, six minute walk tests and radiology reports will be collected as they occur.