Prediction of elevated pulse pressure in elderly using in-home monitoring sensors: A pilot study

In this paper we describe the possibility of employing the data generated by a continuous, unobtrusive home monitoring system for predicting abnormal blood pressure (BP) in elderly. Blood pressure may be used for both early detection of clinical conditions (such as heart attacks or strokes) and long term assessment of functional or cognitive decline. We investigated several factors that influence abnormal BP prediction such as sensor type, number of days prior to the BP measurement and algorithm choice. In our algorithms we used the pulse pressure (the difference between systolic and diastolic BP) that is believed to be a better predictor for clinical events. We conducted a retrospective pilot study on two residents of the TigerPlace aging in place facility with age over 70, that had blood pressure measured between 100 and 300 times during a period of two years. The pilot study suggested that abnormal pulse pressure can be reasonably well estimated (an area under ROC curve of about 0.75) using apartment bed and motion sensors.