Piecewise-linear trend detection in longitudinal physiological measurements

Recently, telecare solutions have been demonstrated as an effective means of monitoring chronic disease at a distance. A clinician may be managing many tens or hundreds of remote patients, prompting the need for a decision support system (DSS) to provide a more automated approach to managing these vast amounts of data. While simple threshold-based alert techniques provide some utility in notifying clinicians of extreme out-of-range parameter values, more incipient changes in a subject´s condition may be sooner recognized by identifying trends in the longitudinal parameter data. Here we describe an approach for obtaining a piecewise-linear fit, to longitudinal physiological trend data, comparable with a similar fitting performed by a human observer, using a graphical user interface. The technique has been applied to both simulated and real data, and a comparison performed against the human scoring for each. On simulated data, the method matches or betters the human performance in most cases; with the greatest improvement observed in more noisy data. Similarly, for real physiological data, the deviation from the human marking, as a fraction of total variability of the signal, is less than 0.35.